~/talk

Build Your Own Data Science AI Agents

on YouTube 01h 29m
Talkat PyData Virginia 2025 · 2025-04-18..04-19· [ai-assisted-development, machine-learning]

speakers

name:Krishnan
links:none on file
name:Liu
links:none on file
name:Puri
links:none on file
name:Rojas
links:none on file

resources

video:https://youtu.be/s5dx_4y6Iy8
slides:none on file

summary

Claim: a runnable multi-agent pipeline that reads diabetes.csv, does EDA, trains a predictive model and auto-generates an executive PPT. Tools actually used: OpenAI GPT‑4o as the LLM, lang chain and langraph for graph nodes, Langsmith for tracing, Streamlit UI, sklearn RandomForest for modeling. Architecture is a supervisor -> analyst -> scientist -> reporter graph with shared state, a recursion limit of 20, and the guideline that about 5 to 7 tools keeps agents stable. Demo mostly completes the workflow though occasional API quota and demo hardware hiccups appear.

AI-written summary · about

chapters (16)

  1. 0:00Introduction
  2. 2:34AI agents and use cases
  3. 6:50AI agent types and frameworks
  4. 10:18Single vs multi-agent systems
  5. 15:18Supervisor agent architecture
  6. 21:59Setting up the environment and dependencies
  7. 33:37Diabetes dataset exploration
  8. 35:29State and supervisor config
  9. 44:00Data analyst agent
  10. 57:31Data scientist agent
  11. 1:02:04Reporter agent
  12. 1:08:24Multi-agent workflow graph
  13. 1:11:53Running the system
  14. 1:19:18Production deployment
  15. 1:22:53Streamlit UI demo
  16. 1:25:22Conclusion

related talks

~/search
type to search 6,249 conferences, 138,987 speakers and 185,172 talks
~/help
cmdctrl k
search every conference, speaker and series
/
focus the filter search
h
home
c
conferences
s
speakers
e
series
t
light / dark — the status bar button does the same
?
this help
esc
close this, then clear every filter — or just the search, while you are typing in it

Ignored while you are typing. Pages without a filter rail have no search to focus.