title
Interview With My Brother Who Sold His Startup For $60 Million | Machine Learning Engineer

description
My brother (https://twitter.com/madavidj) talks about how to successfully leverage Machine Learning in Startups. Learn computer science, math, science, and algorithms at http://brilliant.org/joma (first 200 get 20% off premium). HOW DO I GET A TECH JOB? ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 📚 Video courses from JomaClass: 🎓 New to programming? Learn Python here: https://bit.ly/joma_python 🎓 Learn SQL for data science and data analytics: https://bit.ly/joma_sql 🎓 Data Structures and Algorithms: https://bit.ly/joma_dsa 💼 Resume Template and Cover letter I used for applying to software internships and full-time jobs: https://resume.joma.io 💼 Interviewing for jobs now? Get access to interview question database, courses, coaching, and peer community today: https://www.tryexponent.com/?ref=joma 📱 SOCIAL MEDIA ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ https://www.instagram.com/jomakaze/ https://twitter.com/jomakaze https://www.facebook.com/jomakaze Some of the links in this description are affiliate links that I get a kickback from

detail
{'title': 'Interview With My Brother Who Sold His Startup For $60 Million | Machine Learning Engineer', 'heatmap': [{'end': 640.85, 'start': 591.52, 'weight': 0.781}, {'end': 852.538, 'start': 824.097, 'weight': 1}], 'summary': "Features an interview with david ma, who sold his startup, dynasty, to appfolio for $60 million, discussing the startup's success, machine learning misconceptions, integration challenges, ai chatbot success, ml in chat application, and challenges in hiring and implementing ai.", 'chapters': [{'end': 94.367, 'segs': [{'end': 56.124, 'src': 'embed', 'start': 25.84, 'weight': 0, 'content': [{'end': 26.82, 'text': 'We sold for $60 million.', 'start': 25.84, 'duration': 0.98}, {'end': 31.76, 'text': 'Before we continue this video, I just want to say thank you Brilliant for sponsoring this video.', 'start': 27.216, 'duration': 4.544}, {'end': 39.288, 'text': 'Everyday Brilliant publishes daily challenges on many STEM topics like math, science, and computer science.', 'start': 32.241, 'duration': 7.047}, {'end': 44.473, 'text': 'This site is extremely sleek and they have over 60 interactive courses,', 'start': 39.788, 'duration': 4.685}, {'end': 48.377, 'text': 'which makes learning these concepts way easier because of the hands-on approach.', 'start': 44.473, 'duration': 3.904}, {'end': 56.124, 'text': "They also have an artificial neural networks course, which is really really well made, and I think, if you're very interested in ML,", 'start': 48.936, 'duration': 7.188}], 'summary': 'Sold for $60 million, sponsored by brilliant, offering 60+ interactive courses.', 'duration': 30.284, 'max_score': 25.84, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/fB7nyxXaczY/pics/fB7nyxXaczY25840.jpg'}], 'start': 0.209, 'title': 'Sale to appfolio', 'summary': 'Discusses the $60 million sale of a product to appfolio and a sponsorship message for brilliant, an online platform offering daily stem challenges and over 60 interactive courses.', 'chapters': [{'end': 94.367, 'start': 0.209, 'title': 'Sale of product to appfolio', 'summary': 'Discusses the sale of a product to appfolio for $60 million, as well as a sponsorship message for brilliant, an online platform offering daily stem challenges and over 60 interactive courses.', 'duration': 94.158, 'highlights': ["The product was sold to Appfolio for $60 million, and the speaker is currently involved in scaling and adapting the product to Appfolio's client base.", 'Brilliant, an online platform, is being promoted in the video, offering daily STEM challenges and over 60 interactive courses, including an artificial neural networks course.', "The speaker expresses a wish for having access to Brilliant during college, highlighting the platform's hands-on approach to learning and providing a 20% discount for the first 200 people who sign up using a specific link."]}], 'duration': 94.158, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/fB7nyxXaczY/pics/fB7nyxXaczY209.jpg', 'highlights': ["The product was sold to Appfolio for $60 million, and the speaker is currently involved in scaling and adapting the product to Appfolio's client base.", 'Brilliant, an online platform, is being promoted in the video, offering daily STEM challenges and over 60 interactive courses, including an artificial neural networks course.', "The speaker expresses a wish for having access to Brilliant during college, highlighting the platform's hands-on approach to learning and providing a 20% discount for the first 200 people who sign up using a specific link."]}, {'end': 389.835, 'segs': [{'end': 124.342, 'src': 'embed', 'start': 94.467, 'weight': 0, 'content': [{'end': 95.308, 'text': 'Thanks for having me back.', 'start': 94.467, 'duration': 0.841}, {'end': 97.148, 'text': 'Yeah, thanks for coming.', 'start': 96.268, 'duration': 0.88}, {'end': 98.749, 'text': 'David Ma, my brother.', 'start': 97.308, 'duration': 1.441}, {'end': 103.955, 'text': "Cool I think I'm your first interviewer that came back for a second round.", 'start': 99.873, 'duration': 4.082}, {'end': 106.075, 'text': 'Am I right? I think you are, actually.', 'start': 104.035, 'duration': 2.04}, {'end': 108.276, 'text': 'Yeah Interesting.', 'start': 106.095, 'duration': 2.181}, {'end': 110.837, 'text': 'Because you were high in demand, so I had to bring you back.', 'start': 108.296, 'duration': 2.541}, {'end': 113.598, 'text': 'Cool So just a little context.', 'start': 111.817, 'duration': 1.781}, {'end': 120.06, 'text': "In the previous video, I made an interview with you, and it's mostly about how you were a quant at Two Sigma.", 'start': 113.878, 'duration': 6.182}, {'end': 124.342, 'text': 'And then in that video, you told me that you quit your job.', 'start': 120.461, 'duration': 3.881}], 'summary': 'Interviewer welcomes back david ma, a former quant at two sigma, for a second round interview.', 'duration': 29.875, 'max_score': 94.467, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/fB7nyxXaczY/pics/fB7nyxXaczY94467.jpg'}, {'end': 389.835, 'src': 'embed', 'start': 336.232, 'weight': 1, 'content': [{'end': 343.341, 'text': "I'm not going to say it's a bad idea because we still think that there are benefits to this world we're dreaming of.", 'start': 336.232, 'duration': 7.109}, {'end': 346.023, 'text': 'But in the process,', 'start': 344.522, 'duration': 1.501}, {'end': 355.752, 'text': 'Elliot and other people who joined before me found out that a lot of real estate participants had a lot of trouble managing their assets.', 'start': 346.023, 'duration': 9.729}, {'end': 362.219, 'text': "So it's unlike stocks, real estate is an asset that you have to you know, it's a real thing, you have to..", 'start': 355.873, 'duration': 6.346}, {'end': 363.34, 'text': "There's upkeep.", 'start': 362.219, 'duration': 1.121}, {'end': 367.964, 'text': "Yeah, there's upkeep, you want to get people in for rentals and stuff like that.", 'start': 363.56, 'duration': 4.404}, {'end': 372.008, 'text': 'I think half of the income comes from rentals, right?', 'start': 368.565, 'duration': 3.443}, {'end': 375.351, 'text': 'If you take the other half as appreciation.', 'start': 372.869, 'duration': 2.482}, {'end': 384.294, 'text': 'So one of the big challenges was the operations of leasing a building or leasing your assets.', 'start': 376.172, 'duration': 8.122}, {'end': 388.855, 'text': "So we decided, well, everybody says it's a problem, so let's do something about it.", 'start': 384.834, 'duration': 4.021}, {'end': 389.835, 'text': 'And solve it.', 'start': 389.295, 'duration': 0.54}], 'summary': 'Real estate operations are challenging, with half the income from rentals and half from appreciation. efforts to solve these challenges are underway.', 'duration': 53.603, 'max_score': 336.232, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/fB7nyxXaczY/pics/fB7nyxXaczY336232.jpg'}], 'start': 94.467, 'title': 'Startup success stories', 'summary': "Explores david ma's transition from being a quant at two sigma to co-founding a startup, dynasty, after being approached by the co-founder elliot, achieving product market fit, and selling to appfolio for 60 million, as well as the evolution of the co-founding team from 10 members to 5 during the pivot. it also discusses the shift from securitizing real estate assets to addressing real estate asset management challenges, driven by the high capital requirement for real estate investment and the operational complexities involved.", 'chapters': [{'end': 189.314, 'start': 94.467, 'title': 'David ma: from quant to startup co-founder', 'summary': "Explores david ma's transition from being a quant at two sigma to co-founding a startup, dynasty, after being approached by the co-founder elliot soon after quitting, with interests in deep learning, cryptocurrencies, and biotech.", 'duration': 94.847, 'highlights': ['David Ma transitioned from being a quant at Two Sigma to co-founding Dynasty after being approached by the co-founder Elliot soon after quitting.', "David Ma's interests in exploring deep learning, cryptocurrencies, and biotech motivated his decision to quit and try something new.", 'Elliot, the co-founder of Dynasty, reached out to David a week after he quit, ultimately leading to David joining Dynasty.']}, {'end': 276.464, 'start': 189.514, 'title': 'Ai real estate startup success', 'summary': 'Discusses the successful pivot of a real estate startup into ai, achieving product market fit, and selling to appfolio for 60 million, as well as the evolution of the co-founding team from 10 members to 5 during the pivot.', 'duration': 86.95, 'highlights': ['The startup sold to Appfolio for 60 million. The company successfully sold to Appfolio for a substantial amount, indicating the value of the business in the real estate market.', 'Pivoted into AI for real estate business and found product market fit. The company successfully pivoted into AI for real estate and found product market fit, demonstrating the effectiveness of their strategy in meeting market demands.', "Evolution of co-founding team from 10 members to 5 during the pivot. The co-founding team transitioned from 10 members to 5 during the pivot, indicating a shift in the company's structure and focus."]}, {'end': 389.835, 'start': 283.609, 'title': 'Real estate asset management solution', 'summary': 'Discusses the shift from securitizing real estate assets to addressing real estate asset management challenges, driven by the high capital requirement for real estate investment and the operational complexities involved.', 'duration': 106.226, 'highlights': ['The initial goal was to create an exchange for securitizing real estate assets to enable individuals with smaller capital to invest, aiming to democratize real estate investment. However, the concept did not succeed.', 'Real estate participants faced challenges in managing their assets due to the operational complexities involved, particularly in leasing and upkeep, with approximately half of the income coming from rentals.', 'The shift in focus was driven by the recognition of the difficulties in real estate asset management, leading to the decision to address these challenges and provide a solution.']}], 'duration': 295.368, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/fB7nyxXaczY/pics/fB7nyxXaczY94467.jpg', 'highlights': ['Dynasty co-founder David Ma transitioned from being a quant at Two Sigma to co-founding Dynasty after being approached by co-founder Elliot.', 'Startup sold to Appfolio for 60 million, indicating the value of the business in the real estate market.', 'Pivoted into AI for real estate business and found product market fit, demonstrating the effectiveness of their strategy in meeting market demands.', "Evolution of co-founding team from 10 members to 5 during the pivot, indicating a shift in the company's structure and focus.", 'The shift in focus was driven by the recognition of the difficulties in real estate asset management, leading to the decision to address these challenges and provide a solution.']}, {'end': 715.432, 'segs': [{'end': 470.333, 'src': 'embed', 'start': 414.508, 'weight': 0, 'content': [{'end': 419.471, 'text': 'I think the general excitement about ML is great.', 'start': 414.508, 'duration': 4.963}, {'end': 424.213, 'text': "It made a lot of people go into ML and that's awesome.", 'start': 420.031, 'duration': 4.182}, {'end': 432.298, 'text': 'But a lot of focus has been on how do I build models and how do I fit a model to the data?', 'start': 424.473, 'duration': 7.825}, {'end': 437.24, 'text': 'But very little focus has been on how do I generate data?', 'start': 432.398, 'duration': 4.842}, {'end': 443.484, 'text': 'How do I design a business process that will create data for the algorithm that I want to build?', 'start': 437.761, 'duration': 5.723}, {'end': 445.424, 'text': 'How do I handle the outputs?', 'start': 444.304, 'duration': 1.12}, {'end': 449.346, 'text': 'How do I build all the process around the ML components?', 'start': 445.765, 'duration': 3.581}, {'end': 453.287, 'text': "Yeah, there's too much focus on building the models.", 'start': 449.926, 'duration': 3.361}, {'end': 459.609, 'text': 'not enough focus on how to integrate ML into existing products.', 'start': 453.287, 'duration': 6.322}, {'end': 468.072, 'text': "And to be fair, it's kind of a new field, right? Not many people know how to do this because it's so new.", 'start': 460.169, 'duration': 7.903}, {'end': 470.333, 'text': 'Like an analogy is.', 'start': 468.852, 'duration': 1.481}], 'summary': 'Excitement about ml has led to focus on model building, but little on data generation and integration into existing products.', 'duration': 55.825, 'max_score': 414.508, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/fB7nyxXaczY/pics/fB7nyxXaczY414508.jpg'}, {'end': 533.238, 'src': 'embed', 'start': 498.599, 'weight': 6, 'content': [{'end': 500.22, 'text': "That's the example I want to use.", 'start': 498.599, 'duration': 1.621}, {'end': 503.302, 'text': 'I always wondered, a lot of people want to do machine learning now.', 'start': 500.6, 'duration': 2.702}, {'end': 505.983, 'text': 'My viewers especially.', 'start': 503.642, 'duration': 2.341}, {'end': 509.525, 'text': "I think it's because I'm in the intersection of data science and software engineering.", 'start': 506.343, 'duration': 3.182}, {'end': 515.389, 'text': "But I don't really understand the appeal of machine learning.", 'start': 510.366, 'duration': 5.023}, {'end': 523.073, 'text': 'at the job, because in my mind, what you do is like you said, you make sure you have good data, make sure you solve a problem with your ML.', 'start': 516.369, 'duration': 6.704}, {'end': 525.734, 'text': 'So most of the time in my head.', 'start': 523.393, 'duration': 2.341}, {'end': 533.238, 'text': 'you build data pipelines, you funnel it into your model, you pick a model, you play with the,', 'start': 525.734, 'duration': 7.504}], 'summary': 'Interest in machine learning among viewers, need for good data and problem-solving in ml.', 'duration': 34.639, 'max_score': 498.599, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/fB7nyxXaczY/pics/fB7nyxXaczY498599.jpg'}, {'end': 620.202, 'src': 'embed', 'start': 591.52, 'weight': 2, 'content': [{'end': 603.716, 'text': "What would be fun for me, I think, in terms of research, would be investigating the latest algorithms and understanding why they work, why they don't.", 'start': 591.52, 'duration': 12.196}, {'end': 608.198, 'text': 'trying different data sets on these new algorithms.', 'start': 604.497, 'duration': 3.701}, {'end': 618.181, 'text': "A lot of the things that you've seen out there, like GANs, like generative adversarial networks, you make like funky images, style transfers.", 'start': 608.538, 'duration': 9.643}, {'end': 620.202, 'text': 'Like these were all investigations.', 'start': 618.421, 'duration': 1.781}], 'summary': 'Researching latest algorithms, testing on new data sets, exploring gans and style transfers.', 'duration': 28.682, 'max_score': 591.52, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/fB7nyxXaczY/pics/fB7nyxXaczY591520.jpg'}, {'end': 640.85, 'src': 'heatmap', 'start': 591.52, 'weight': 0.781, 'content': [{'end': 603.716, 'text': "What would be fun for me, I think, in terms of research, would be investigating the latest algorithms and understanding why they work, why they don't.", 'start': 591.52, 'duration': 12.196}, {'end': 608.198, 'text': 'trying different data sets on these new algorithms.', 'start': 604.497, 'duration': 3.701}, {'end': 618.181, 'text': "A lot of the things that you've seen out there, like GANs, like generative adversarial networks, you make like funky images, style transfers.", 'start': 608.538, 'duration': 9.643}, {'end': 620.202, 'text': 'Like these were all investigations.', 'start': 618.421, 'duration': 1.781}, {'end': 624.843, 'text': 'and why do convolutional networks work as they do?', 'start': 620.202, 'duration': 4.641}, {'end': 627.344, 'text': "And that's research, right?", 'start': 626.184, 'duration': 1.16}, {'end': 628.404, 'text': "And that's not.", 'start': 627.424, 'duration': 0.98}, {'end': 628.724, 'text': "it's not.", 'start': 628.404, 'duration': 0.32}, {'end': 632.706, 'text': 'primarily, those things were not built primarily for a business.', 'start': 628.724, 'duration': 3.982}, {'end': 633.506, 'text': 'Got it.', 'start': 632.926, 'duration': 0.58}, {'end': 640.85, 'text': 'And of course, Lisa is an AI for leasing, the pun.', 'start': 635.908, 'duration': 4.942}], 'summary': 'Researching latest algorithms, testing on new data sets, investigating gans and convolutional networks for ai development.', 'duration': 49.33, 'max_score': 591.52, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/fB7nyxXaczY/pics/fB7nyxXaczY591520.jpg'}], 'start': 390.175, 'title': 'Ml misconceptions and integration challenges, and machine learning and ml engineering', 'summary': 'Discusses misconceptions in ml focusing on model building rather than data generation, and the integration challenges, emphasizing the need for a broader understanding of ml engineering. it also highlights the appeal of machine learning, the distinction between ml engineering and ml research, core skills, processes, fun aspects, latest algorithms, and real-world applications in leasing automation.', 'chapters': [{'end': 496.958, 'start': 390.175, 'title': 'Ml misconceptions and integration challenges', 'summary': 'Highlights the misconception of focusing on building models rather than generating data and integrating ml into existing products, emphasizing the need for a broader understanding of ml engineering.', 'duration': 106.783, 'highlights': ['The focus on building models in ML engineering has led to neglect in understanding how to generate data and integrate ML into existing products, posing a challenge in handling outputs and designing business processes (quantifiable data: emphasis on the need for a broader understanding of ML engineering).', 'The excitement around ML has drawn many people into the field, but there is a lack of emphasis on designing business processes to create data for algorithms, indicating the need for a shift in focus in the field (quantifiable data: emphasis on the need for a shift in focus in the field).', 'The analogy of integrating ML into existing products being similar to the challenges faced during the introduction of computers and the internet into business processes, highlighting the newness of the field and the necessity for trial and error in integrating ML technology (quantifiable data: emphasis on the newness of the field and the necessity for trial and error).']}, {'end': 715.432, 'start': 498.599, 'title': 'Machine learning and ml engineering', 'summary': 'Discusses the appeal of machine learning and the distinction between ml engineering and ml research, highlighting the core skills and processes involved, as well as the fun aspects, latest algorithms, and real-world application in leasing automation.', 'duration': 216.833, 'highlights': ['The chapter discusses the appeal of machine learning and the distinction between ML engineering and ML research. It explores the intersection of data science and software engineering, addressing the appeal of machine learning to a wide audience.', 'The core skills and processes involved in ML engineering are explained, emphasizing building data pipelines, model selection, parameter tuning, and optimization for AUC. The chapter delves into the core activities of ML engineering, focusing on the importance of good data, model selection, and parameter tuning for AUC optimization.', 'The fun aspects of ML research are highlighted, including investigating the latest algorithms, understanding their functionality, and experimenting with different datasets. It emphasizes the enjoyable aspects of ML research, such as exploring new algorithms, understanding their workings, and experimenting with diverse datasets.', 'Real-world application in leasing automation is discussed, demonstrating the practical use of ML in automating responses to text messages, emails, and phone calls for property showings. The chapter illustrates the application of ML in leasing automation, specifically in automating responses to communication channels and organizing property showings, resulting in reduced coordination efforts and positive user feedback.']}], 'duration': 325.257, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/fB7nyxXaczY/pics/fB7nyxXaczY390175.jpg', 'highlights': ['The focus on building models in ML engineering has led to neglect in understanding how to generate data and integrate ML into existing products, posing a challenge in handling outputs and designing business processes.', 'The excitement around ML has drawn many people into the field, but there is a lack of emphasis on designing business processes to create data for algorithms, indicating the need for a shift in focus in the field.', 'The analogy of integrating ML into existing products being similar to the challenges faced during the introduction of computers and the internet into business processes, highlighting the newness of the field and the necessity for trial and error in integrating ML technology.', 'The chapter discusses the appeal of machine learning and the distinction between ML engineering and ML research. It explores the intersection of data science and software engineering, addressing the appeal of machine learning to a wide audience.', 'The core skills and processes involved in ML engineering are explained, emphasizing building data pipelines, model selection, parameter tuning, and optimization for AUC.', 'The fun aspects of ML research are highlighted, including investigating the latest algorithms, understanding their functionality, and experimenting with different datasets.', 'Real-world application in leasing automation is discussed, demonstrating the practical use of ML in automating responses to text messages, emails, and phone calls for property showings.']}, {'end': 1169.9, 'segs': [{'end': 776.917, 'src': 'embed', 'start': 744.557, 'weight': 5, 'content': [{'end': 752.282, 'text': 'build out the ML components, got into deep learning, learned about NLP, which is natural language processing.', 'start': 744.557, 'duration': 7.725}, {'end': 758.285, 'text': 'And once that core thing was built, we were still a startup.', 'start': 752.982, 'duration': 5.303}, {'end': 759.766, 'text': 'We just had to do other things.', 'start': 758.565, 'duration': 1.201}, {'end': 763.068, 'text': 'So I got myself into software engineering.', 'start': 759.906, 'duration': 3.162}, {'end': 766.57, 'text': 'Before that, I had never done real software engineering.', 'start': 764.109, 'duration': 2.461}, {'end': 776.917, 'text': 'But under our CTO I was able to learn a lot about, um, you know how to build.', 'start': 767.371, 'duration': 9.546}], 'summary': 'Transitioned from ml to software engineering in startup, learning nlp and deep learning.', 'duration': 32.36, 'max_score': 744.557, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/fB7nyxXaczY/pics/fB7nyxXaczY744557.jpg'}, {'end': 852.538, 'src': 'heatmap', 'start': 824.097, 'weight': 1, 'content': [{'end': 828.919, 'text': 'About 40% of our messages are handled by humans, who we call operators.', 'start': 824.097, 'duration': 4.822}, {'end': 840.448, 'text': 'The fact that we try to give such a natural experience to Prospect is something that the clients really liked, because, in general,', 'start': 829.759, 'duration': 10.689}, {'end': 842.229, 'text': "people don't like to interact with the bot.", 'start': 840.448, 'duration': 1.781}, {'end': 852.538, 'text': "So our conversations look very natural because there's a lot of humans that we fall back to when things go wrong.", 'start': 843.951, 'duration': 8.587}], 'summary': 'About 40% of messages handled by humans, providing a natural experience favored by clients.', 'duration': 28.441, 'max_score': 824.097, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/fB7nyxXaczY/pics/fB7nyxXaczY824097.jpg'}, {'end': 910.829, 'src': 'embed', 'start': 881.446, 'weight': 3, 'content': [{'end': 891.374, 'text': 'and then people were probably freaked out and replied less than if we were to wait two minutes before replying.', 'start': 881.446, 'duration': 9.928}, {'end': 897.399, 'text': 'So although we can reply very quickly, sometimes we wait a minute or two before doing so.', 'start': 892.034, 'duration': 5.365}, {'end': 898.78, 'text': "That's interesting.", 'start': 897.419, 'duration': 1.361}, {'end': 903.323, 'text': 'So another part that made us successful, I think, was the fact.', 'start': 899.28, 'duration': 4.043}, {'end': 910.829, 'text': 'well, I think the biggest success was the market validation that was done prior to the pivot.', 'start': 903.323, 'duration': 7.506}], 'summary': 'People replied less when there was a two-minute delay in response, contributing to our success in market validation.', 'duration': 29.383, 'max_score': 881.446, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/fB7nyxXaczY/pics/fB7nyxXaczY881446.jpg'}, {'end': 1028.353, 'src': 'embed', 'start': 996.273, 'weight': 2, 'content': [{'end': 998.875, 'text': 'but what did you work on instead to make your product better??', 'start': 996.273, 'duration': 2.602}, {'end': 1006.901, 'text': 'In the backend, the first ML component that we created was an intent classifier.', 'start': 999.336, 'duration': 7.565}, {'end': 1012.885, 'text': 'So we would take in messages and understand what is the intent.', 'start': 1007.041, 'duration': 5.844}, {'end': 1017.767, 'text': 'Classify them as one of a few intents.', 'start': 1012.985, 'duration': 4.782}, {'end': 1021.99, 'text': 'Like, do they want a showing? Do they accept something? Like, whatever.', 'start': 1017.807, 'duration': 4.183}, {'end': 1028.353, 'text': 'Actually, before we continue, can I just kind of have a high-level overview of what Lisa does?', 'start': 1022.01, 'duration': 6.343}], 'summary': 'Developed an intent classifier in the backend to categorize user messages for product improvement.', 'duration': 32.08, 'max_score': 996.273, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/fB7nyxXaczY/pics/fB7nyxXaczY996273.jpg'}, {'end': 1169.9, 'src': 'embed', 'start': 1130.452, 'weight': 0, 'content': [{'end': 1141.319, 'text': 'And also they have a concept called the quick action, which they can quickly find a commonly used action to reply to prospects.', 'start': 1130.452, 'duration': 10.867}, {'end': 1143.06, 'text': 'Awesome Okay.', 'start': 1142.08, 'duration': 0.98}, {'end': 1147.624, 'text': 'So that means back then real estate agent will have to communicate with the prospects.', 'start': 1143.261, 'duration': 4.363}, {'end': 1149.585, 'text': 'do the showing schedule their own showing?', 'start': 1147.624, 'duration': 1.961}, {'end': 1150.586, 'text': 'manage their own calendar?', 'start': 1149.585, 'duration': 1.001}, {'end': 1157.291, 'text': "but now they actually don't even need to talk to the prospects and you just have Lisa that is kind of like a layer in between them.", 'start': 1151.286, 'duration': 6.005}, {'end': 1158.231, 'text': "Oh, that's really cool.", 'start': 1157.431, 'duration': 0.8}, {'end': 1163.735, 'text': 'Until they get to the showing, which they want to use their human specialty to do the sale.', 'start': 1158.291, 'duration': 5.444}, {'end': 1166.718, 'text': 'Well, I mean, one day you could build robots and just replace that too.', 'start': 1163.775, 'duration': 2.943}, {'end': 1169.24, 'text': 'One day, one day.', 'start': 1168.059, 'duration': 1.181}, {'end': 1169.9, 'text': 'Cool Okay.', 'start': 1169.26, 'duration': 0.64}], 'summary': 'Real estate agents can use quick action to reply to prospects, reducing direct communication needs.', 'duration': 39.448, 'max_score': 1130.452, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/fB7nyxXaczY/pics/fB7nyxXaczY1130452.jpg'}], 'start': 717.092, 'title': 'Transition to ml and ai chatbot success', 'summary': "Covers the speaker's transition to ml and software engineering, including deep learning and nlp, and discusses the success of an ai leasing chatbot, achieving a 40% human message handling rate and emphasizing market validation and simplicity over technology optimization.", 'chapters': [{'end': 798.734, 'start': 717.092, 'title': 'Transition to ml and software engineering', 'summary': "Describes the speaker's transition from research to ml and software engineering, including deep learning, nlp, and product design at a startup.", 'duration': 81.642, 'highlights': ["The speaker dived into ML and deep learning, transitioning from a research role, and later ventured into NLP. (Relevant for describing the speaker's transition to ML and deep learning)", "The speaker also acquired skills in software engineering and product design under the guidance of the CTO at a startup. (Relevant for illustrating the speaker's expansion into software engineering and product design)", "The speaker's experience reflects the necessity of being versatile at a startup, as they had to engage in various tasks including AI application to real estate. (Relevant for highlighting the speaker's adaptability and diverse responsibilities at a startup)"]}, {'end': 1169.9, 'start': 800.208, 'title': 'Success with ai leasing chatbot', 'summary': "Discusses the success of an ai leasing chatbot, highlighting that about 40% of messages are handled by humans to provide a natural experience, achieving a higher reply ratio by delaying responses, and focusing on market validation and simplicity over technology optimization, leading to the product's success.", 'duration': 369.692, 'highlights': ["About 40% of messages are handled by humans to provide a natural experience to prospects, which clients appreciate. The company's approach involves using humans to handle about 40% of messages to offer a natural experience to prospects, a strategy well-received by clients.", 'Achieving a higher reply ratio by delaying responses, with people replying less when messages were sent instantaneously. The company observed a higher reply ratio when delaying responses, indicating that people were less inclined to reply when messages were sent instantaneously.', "Focusing on market validation and simplicity over technology optimization contributed to the product's success. The emphasis on market validation and simplicity over technology optimization proved to be a key factor in the product's success, allowing the company to iterate and build the product efficiently."]}], 'duration': 452.808, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/fB7nyxXaczY/pics/fB7nyxXaczY717092.jpg', 'highlights': ["The emphasis on market validation and simplicity over technology optimization proved to be a key factor in the product's success.", "The speaker's experience reflects the necessity of being versatile at a startup, as they had to engage in various tasks including AI application to real estate.", 'The speaker also acquired skills in software engineering and product design under the guidance of the CTO at a startup.', 'The speaker dived into ML and deep learning, transitioning from a research role, and later ventured into NLP.', 'Achieving a higher reply ratio by delaying responses, with people replying less when messages were sent instantaneously.', 'About 40% of messages are handled by humans to provide a natural experience to prospects, which clients appreciate.']}, {'end': 1487.095, 'segs': [{'end': 1220.992, 'src': 'embed', 'start': 1192.678, 'weight': 4, 'content': [{'end': 1200.762, 'text': 'Once we got the intent classifiers, we could have, you know, tried better models like BERT or ELMO,', 'start': 1192.678, 'duration': 8.084}, {'end': 1203.823, 'text': 'like the stuff that came out in 2008 and was really hot.', 'start': 1200.762, 'duration': 3.061}, {'end': 1209.706, 'text': 'We could have tried using that to, you know, gain a few percentage points of accuracy.', 'start': 1204.144, 'duration': 5.562}, {'end': 1213.008, 'text': "Are those models like NLP models? That's right, yeah.", 'start': 1209.986, 'duration': 3.022}, {'end': 1220.992, 'text': 'We used something like pretty simple, like a tech CNN for our core models and it worked fine.', 'start': 1213.208, 'duration': 7.784}], 'summary': 'Utilized intent classifiers and a cnn model for nlp, could have explored bert or elmo for improved accuracy.', 'duration': 28.314, 'max_score': 1192.678, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/fB7nyxXaczY/pics/fB7nyxXaczY1192678.jpg'}, {'end': 1284.316, 'src': 'embed', 'start': 1256.344, 'weight': 2, 'content': [{'end': 1262.671, 'text': 'So what do you think most startups make as mistakes using ML??', 'start': 1256.344, 'duration': 6.327}, {'end': 1266.634, 'text': 'Like, what are the common mistakes startups make using ML??', 'start': 1262.871, 'duration': 3.763}, {'end': 1271.177, 'text': "Because I'm guessing you're comparing your dynasty to other startups.", 'start': 1266.754, 'duration': 4.423}, {'end': 1272.958, 'text': 'So how would they do?', 'start': 1271.357, 'duration': 1.601}, {'end': 1273.839, 'text': 'what would they do wrong?', 'start': 1272.958, 'duration': 0.881}, {'end': 1284.316, 'text': "My guess would be focusing too much on the latest technology that's out there, especially the academic literature,", 'start': 1274.426, 'duration': 9.89}], 'summary': 'Startups often make the mistake of focusing too much on the latest technology and academic literature when using ml.', 'duration': 27.972, 'max_score': 1256.344, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/fB7nyxXaczY/pics/fB7nyxXaczY1256344.jpg'}, {'end': 1331.96, 'src': 'embed', 'start': 1308.687, 'weight': 0, 'content': [{'end': 1316.552, 'text': 'i heard a lot of people saying that in theory, the paper sounds great and it works with their data set, but applying it is a whole different story.', 'start': 1308.687, 'duration': 7.865}, {'end': 1318.213, 'text': 'yeah, applying you have to.', 'start': 1316.552, 'duration': 1.661}, {'end': 1331.06, 'text': 'so finding a way to apply ml to a business environment is difficult because you have to specifically know which problem like business in itself is like many,', 'start': 1318.213, 'duration': 12.847}, {'end': 1331.96, 'text': 'many problems.', 'start': 1331.06, 'duration': 0.9}], 'summary': 'Challenges in applying ml to business due to diverse problems and data sets.', 'duration': 23.273, 'max_score': 1308.687, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/fB7nyxXaczY/pics/fB7nyxXaczY1308687.jpg'}], 'start': 1170.16, 'title': 'Ml in chat application and common mistakes in startups using ml', 'summary': 'Discusses the use of machine learning (ml) in a chat application, including intent classifiers and prioritizing ml solutions in other parts of the business. it also covers common mistakes startups make with ml, such as focusing too much on latest technology and not understanding specific business needs.', 'chapters': [{'end': 1254.102, 'start': 1170.16, 'title': 'Ml in chat application', 'summary': 'Discusses the use of machine learning (ml) in a chat application, including the use of intent classifiers and the decision to prioritize packaging ml solutions in other parts of the business over optimizing existing components.', 'duration': 83.942, 'highlights': ['The decision to prioritize packaging ML solutions in other parts of the business over optimizing existing components was made after implementing intent classifiers.', 'The possibility of using advanced models like BERT or ELMO to gain a few percentage points of accuracy in the ML components was considered.', 'The use of a tech CNN for core models was found to work fine in the chat application.']}, {'end': 1487.095, 'start': 1256.344, 'title': 'Common mistakes in startups using ml', 'summary': 'Discusses common mistakes startups make when using ml, including focusing too much on the latest technology, trying to solve too many problems with ml, and not understanding the intricacies of applying ml to specific business needs.', 'duration': 230.751, 'highlights': ['Focusing too much on the latest technology and academic literature can waste a lot of time and is a common mistake for startups using ML. Startups often make the mistake of investing too much time in the latest technology and academic literature, which may not be reproducible for business, leading to a waste of time.', 'Trying to solve too many problems with ML can be a pitfall for startups, as it is difficult to apply ML to a business environment without understanding the specific problem that needs to be solved. Startups may struggle with trying to solve too many problems with ML, without specifically knowing which business problem to address, leading to inefficiencies and wasted resources.', 'Not understanding the intricacies of applying ML to specific business needs, such as monitoring model performance and handling false positives and false negatives, is a common challenge for startups. Startups often face challenges in applying ML to specific business needs, including monitoring model performance and handling errors like false positives and false negatives, which require a thorough understanding of the intricacies involved.']}], 'duration': 316.935, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/fB7nyxXaczY/pics/fB7nyxXaczY1170160.jpg', 'highlights': ['The decision to prioritize packaging ML solutions in other parts of the business over optimizing existing components was made after implementing intent classifiers.', 'The use of a tech CNN for core models was found to work fine in the chat application.', 'The possibility of using advanced models like BERT or ELMO to gain a few percentage points of accuracy in the ML components was considered.', 'Focusing too much on the latest technology and academic literature can waste a lot of time and is a common mistake for startups using ML.', 'Trying to solve too many problems with ML can be a pitfall for startups, as it is difficult to apply ML to a business environment without understanding the specific problem that needs to be solved.', 'Not understanding the intricacies of applying ML to specific business needs, such as monitoring model performance and handling false positives and false negatives, is a common challenge for startups.']}, {'end': 1758.761, 'segs': [{'end': 1553.072, 'src': 'embed', 'start': 1527.203, 'weight': 1, 'content': [{'end': 1536.331, 'text': "If it's really repetitive and people don't make a lot of mistakes, write some heuristics, not even ML, and run with it for a while.", 'start': 1527.203, 'duration': 9.128}, {'end': 1540.334, 'text': 'Handle your false positives, handle your false negatives.', 'start': 1537.092, 'duration': 3.242}, {'end': 1548.369, 'text': 'And then if the system is humming, then try to increase the accuracy with ML.', 'start': 1541.465, 'duration': 6.904}, {'end': 1553.072, 'text': "But without these intermediate steps, don't even think about it.", 'start': 1549.43, 'duration': 3.642}], 'summary': 'Start with heuristics, handle false positives and negatives, then improve accuracy with ml.', 'duration': 25.869, 'max_score': 1527.203, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/fB7nyxXaczY/pics/fB7nyxXaczY1527203.jpg'}, {'end': 1758.761, 'src': 'embed', 'start': 1700.874, 'weight': 0, 'content': [{'end': 1706.217, 'text': "And after that, who knows? I think I'll be at Appfolio for the foreseeable future.", 'start': 1700.874, 'duration': 5.343}, {'end': 1714.201, 'text': "Maybe I'll be building other ML products, hopefully, finding other ways to apply ML into the real world.", 'start': 1707.838, 'duration': 6.363}, {'end': 1715.842, 'text': 'Awesome Cool.', 'start': 1714.961, 'duration': 0.881}, {'end': 1717.242, 'text': 'Yeah, thank you so much.', 'start': 1716.382, 'duration': 0.86}, {'end': 1722.865, 'text': 'And I just want to say best of luck to Lisa, Dynasty, and Appfolio.', 'start': 1717.543, 'duration': 5.322}, {'end': 1724.906, 'text': 'If they do want to apply to..', 'start': 1723.806, 'duration': 1.1}, {'end': 1734.82, 'text': 'Do they have to do it through Appfolio website or is there a separate Dynasty website? We are fully under Appfolio now.', 'start': 1726.187, 'duration': 8.633}, {'end': 1737.122, 'text': 'So you should apply to Appfolio.', 'start': 1735.44, 'duration': 1.682}, {'end': 1742.106, 'text': "Exactly And then if you want to prepare for Appfolio, don't forget to check out Tech Interview Pro.", 'start': 1737.682, 'duration': 4.424}, {'end': 1746.109, 'text': "Check out Tech Interview Pro if you're interested in getting ready for interviews.", 'start': 1742.787, 'duration': 3.322}, {'end': 1746.53, 'text': 'Oh, yeah.', 'start': 1746.27, 'duration': 0.26}, {'end': 1749.552, 'text': 'I also want to plug my Twitter, MADavidJ.', 'start': 1746.55, 'duration': 3.002}, {'end': 1752.755, 'text': 'MADavidJ Shitposts.', 'start': 1750.553, 'duration': 2.202}, {'end': 1754.016, 'text': 'Cool All right.', 'start': 1753.096, 'duration': 0.92}, {'end': 1756.058, 'text': 'Thank you so much for being here.', 'start': 1754.056, 'duration': 2.002}, {'end': 1757.139, 'text': 'Really appreciate it.', 'start': 1756.238, 'duration': 0.901}, {'end': 1758.2, 'text': 'Thank you for having me.', 'start': 1757.159, 'duration': 1.041}, {'end': 1758.761, 'text': 'All right.', 'start': 1758.22, 'duration': 0.541}], 'summary': 'The speaker plans to stay at appfolio, build ml products, and wishes luck to colleagues. also, recommends tech interview pro and promotes twitter handle madavidj.', 'duration': 57.887, 'max_score': 1700.874, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/fB7nyxXaczY/pics/fB7nyxXaczY1700874.jpg'}], 'start': 1488.135, 'title': 'Challenges in hiring and implementing ai', 'summary': "Addresses challenges in hiring a personal assistant, implementing machine learning, emphasizing the importance of manual problem-solving before considering ai, and the need to handle false positives and false negatives. it also discusses dynasty's current hiring needs, technology stack, and future plans for ml product development and application.", 'chapters': [{'end': 1553.072, 'start': 1488.135, 'title': 'Challenges in hiring and implementing ai', 'summary': 'Addresses the challenge of hiring a personal assistant and the guidelines for implementing machine learning, emphasizing the importance of solving problems manually before considering ai, and the need to handle false positives and false negatives.', 'duration': 64.937, 'highlights': ['The challenge of hiring a personal assistant is still very hard, making it difficult to consider implementing a machine learning model (relevance: 5)', 'Guidelines for implementing machine learning suggest solving problems manually first, evaluating repetition and human error, using heuristics before ML, and handling false positives and false negatives (relevance: 4)', 'The importance of solving problems manually before considering AI, with the suggestion to use heuristics and handle false positives and false negatives before implementing machine learning (relevance: 3)']}, {'end': 1758.761, 'start': 1553.933, 'title': 'Dynasty hiring and future plans', 'summary': "Discusses dynasty's current hiring needs for software engineers, technology stack, and the importance of general software engineering skills for ml engineers, as well as the future plans for lisa and appfolio, emphasizing the potential for ml product development and application. the interviewee also mentions the merger of dynasty and appfolio, and directs potential applicants to apply through appfolio's website.", 'duration': 204.828, 'highlights': ['Dynasty is actively hiring and requires a lot of good engineers with decent software engineering experience, particularly emphasizing the importance of general software engineering skills for ML engineers.', 'The technology stack at Dynasty includes TensorFlow, Python packages for applied machine learning, and Java for backend development, priding themselves on using tried and proven technologies.', 'The interviewee expresses a vision for the future, aiming to see Lisa built out to its full potential, potentially exploring other ML product development opportunities and ways to apply ML into the real world.', "The interviewee also mentions the merger of Dynasty and Appfolio, directing potential applicants to apply through Appfolio's website and suggests Tech Interview Pro for interview preparation, and promotes his Twitter account, MADavidJ."]}], 'duration': 270.626, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/fB7nyxXaczY/pics/fB7nyxXaczY1488135.jpg', 'highlights': ['The challenge of hiring a personal assistant is still very hard, making it difficult to consider implementing a machine learning model (relevance: 5)', 'Guidelines for implementing machine learning suggest solving problems manually first, evaluating repetition and human error, using heuristics before ML, and handling false positives and false negatives (relevance: 4)', 'Dynasty is actively hiring and requires a lot of good engineers with decent software engineering experience, particularly emphasizing the importance of general software engineering skills for ML engineers.', 'The technology stack at Dynasty includes TensorFlow, Python packages for applied machine learning, and Java for backend development, priding themselves on using tried and proven technologies.', 'The interviewee expresses a vision for the future, aiming to see Lisa built out to its full potential, potentially exploring other ML product development opportunities and ways to apply ML into the real world.']}], 'highlights': ["The product was sold to Appfolio for $60 million, and the speaker is currently involved in scaling and adapting the product to Appfolio's client base.", 'Startup sold to Appfolio for 60 million, indicating the value of the business in the real estate market.', 'Pivoted into AI for real estate business and found product market fit, demonstrating the effectiveness of their strategy in meeting market demands.', "The emphasis on market validation and simplicity over technology optimization proved to be a key factor in the product's success.", 'Real-world application in leasing automation is discussed, demonstrating the practical use of ML in automating responses to text messages, emails, and phone calls for property showings.', 'The decision to prioritize packaging ML solutions in other parts of the business over optimizing existing components was made after implementing intent classifiers.', 'The use of a tech CNN for core models was found to work fine in the chat application.', 'The possibility of using advanced models like BERT or ELMO to gain a few percentage points of accuracy in the ML components was considered.', 'The challenge of hiring a personal assistant is still very hard, making it difficult to consider implementing a machine learning model (relevance: 5)', 'Guidelines for implementing machine learning suggest solving problems manually first, evaluating repetition and human error, using heuristics before ML, and handling false positives and false negatives (relevance: 4)']}