Schedule
- AgentWeave: Exploring Multi-Agent Reasoning with Python
- Building a virtual transistor using python
- LLM Optimizations for Local Inference
Lightning talks
- Open Slots
Summary:
AgentWeave: Exploring Multi-Agent Reasoning with Python
Speaker: Kayalvizhi
Do multiple LLM agents improve reasoning, or do they eventually start repeating and agreeing with each other?
AgentWeave is an experimental multi-agent system I built to explore the balance between exploration and convergence.
I'll cover Python-based agent orchestration, contribution evaluation, dynamic agent replacement, and a live demo.
Building a virtual transistor using python
Speaker: Rengaraj
This talk is to simulate a transistor behaviour and build a web application around it using FastAPI. I like to give this talk ground up from understanding math behind transistor operation and using that we are gonna build a transistor in python.
LLM Optimizations for Local Inference
Speaker: Vijay kumar B
Focus on three specific optimizations: NVFP4, speculative decoding, and prefix caching. The talk will begin with a brief introduction to how LLMs work, followed by an explanation of these optimizations, how they work, and how they help accelerate local inference on machines like the DGX Spark.
Talk at Chennaipy
We have three 20 mins talk slots. The talks are filled by first come first serve basis. Lightning talks will be at the end of the meetup. As long as we have time, there is no limit for number of lightning talks.
If you are interested in doing a talk (20 min) or lightning talk (5 mins), please add a comment to this event page with your topic and Description.
Note:
We understand that your schedule might change and you are unable to attend the meetup, in that case consider UNRSVP. So that folks in waiting list will get a chance to attend the meetup.
Talk Submission Reference:
[1]. https://mail.python.org/archives/list/chennaipy@python.org/message/FNKPVXDRKGUYCTSA2LYW464IWVMNDCS4/
