Building Python based AI Systems with LLMs
summary
Claim: LLMs reliably translate text queries into runnable Python, using Hopsworks feature store and pandas to produce feature, training, and inference pipelines. Examples include synthetic credit-card datasets auto-generated and saved to feature groups, a Julia-like CSV-to-plot service that self-corrects parsing errors, and function-calling to fetch live data. Caveats: SQL generation lags, prompts must include API versions and metadata, and automated tests are required to detect model or API drift.
AI-written summary · about
chapters (11)
- 0:00Introduction to AI systems
- 9:32LLMs in programming
- 16:47LLMs as code interfaces
- 20:21Python for LLM code
- 23:53Building AI systems with LLMs
- 27:04Strategic prompting
- 34:42Synthetic data generation
- 37:30Modular AI pipelines
- 41:24LLM development challenges
- 45:41Future of programming with LLMs
- 47:04Q&A
related talks
18mGenerative AI in Financial Industry
38mQuerying Live Data With LLM App
27mAI Demystified Unveiling the Intelligence Behind the Machine
36mLLMs and AI agents demystified
58mAI-Enhanced Big Data: Integrating Private LLMs and Vector Databases
46mBeyond Basic Prompting: Supercharging Open Source LLMs with LMQL's Structured Generation
01h 34mA Practical Starter's Guide to building LLM based projects
38mBuild a personalized Commute virtual assistant in Python w/ Hopsworks & LLM Function Calling