Using Large Language Models at AngelList // Thibaut Labarre // MLOps Podcast #171
MLOps.community
Culture at AngelList and Introduction to Angelisk Relay
The chapter explores the culture at AngelList, including their support for employees starting their own ventures. It also introduces Angelisk Relay, a tool for investors to track and analyze investments.
MLOps Coffee Sessions #171 with Thibaut Labarre, Using Large Language Models at AngelList co-hosted by Ryan Russon. We are now accepting talk proposals for our next LLM in Production virtual conference on October 3rd. Apply to speak here: https://go.mlops.community/NSAX1O // Abstract Thibaut innovatively addressed previous system constraints, achieving scalability and cost efficiency. Leveraging AngelList investing and natural language processing expertise, they refined news article classification for investor dashboards. Central is their groundbreaking platform, AngelList Relay, automating parsing and offering vital insights to investors. Amid challenges like Azure OpenAI collaboration and rate limit solutions, Thibaut reflects candidly. The narrative highlights prompt engineering's strategic importance and empowering domain experts for ongoing advancement. // Bio Thibaut LaBarre is an engineering lead with a background in Natural Language Processing (NLP). Currently, Thibaut focuses on unlocking the potential of Large Language Model (LLM) technology at AngelList, enabling everyone within the organization to become prompt engineers on a quest to streamline and automate the infrastructure for Venture Capital. Prior to that, Thibaut began his journey at Amazon as an intern where he built Heartbeat, a state-of-the-art NLP tool that consolidates millions of data points from various feedback sources, such as product reviews, customer contacts, and social media, to provide valuable insights to global product teams. Over the span of seven years, he expanded his internship project into an organization of 20 engineers. He received a M.S. in Computational Linguistics from the University of Washington. // MLOps Jobs board https://mlops.pallet.xyz/jobs // MLOps Swag/Merch https://mlops-community.myshopify.com/ // Related Links Website: https://www.angellist.com/venture/relay --------------- ✌️Connect With Us ✌️ ------------- Join our slack community: https://go.mlops.community/slack Follow us on Twitter: @mlopscommunity Sign up for the next meetup: https://go.mlops.community/register Catch all episodes, blogs, newsletters, and more: https://mlops.community/ Connect with Demetrios on LinkedIn: https://www.linkedin.com/in/dpbrinkm/ Connect with Ryan on LinkedIn: https://www.linkedin.com/in/ryanrusson/ Connect with Thibaut on LinkedIn: https://www.linkedin.com/in/thibautlabarre/
Timestamps: [00:00] Thibaut's preferred beverage [00:50] Takeaways [04:05] Please like, share, and subscribe to our MLOps channels! [04:44] A huge fan of Isaac Asimov [07:20] Thibaut Labarre background [09:13] AngelList as an organization [10:50] AI sense of building [12:29] System trade-offs [15:20] OpenAI's limitation [16:31] Human in the loop [17:22] Classifying relevance [18:09] Fight for value [19:37] Added value [22:10] Exploring efficient ways to automate tasks. [24:20] Investing in off-the-shelf models [27:56] AngelList Relay [30:49] News article and investment document classification technology [32:39] Back-end tech [34:09] Prompt layer [35:28] Prompt layer as a living [37:04] Foreseeing no human intervention [39:00] Blocking hallucinations [40:33] Challenges [43:49] Investments in other models besides OpenAI [45:20] Integration with other models [46:28] Ethical concerns when [48:37] OpenAI breaking Prompts [50:46] Wrap up
The AI-powered Podcast Player
