Unlocking LLM Potential: From Exploration to Integration
summary
Built a production chatbot called Plato and an experimental customer bot launched in October that uses RAG with embeddings and vector stores (Redis Postgres Mongo Chroma) to answer from internal docs, reducing simple-support turnaround to minutes and improving onboarding and runbook access. Uses LangChain, OpenAI and tries local Llama Mistral deployments, balances cost versus quality, and runs a PR review agent that rates effort 1–5 and emits inline suggestions. Notes 45,000 models on Hugging Face, RAG cuts but does not eliminate hallucination.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 3:30The Journey
- 6:49Lchain
- 17:02Answer
- 17:49Experimenting internally
- 24:35Improving developer experience
- 31:41Learnings
- 32:01Prompt Engineering
- 33:27Security
- 33:53Cloud Providers
- 34:25Bird
- 35:00Small talk
- 35:34Fine tuning
- 36:03Improving the pipeline
- 37:24Chunk size
- 40:46Conclusion
- 41:17Questions
related talks
33mOpening Keynote: LLMs, between fantasy and reality - the poin...
31mSponsored Session: Building a Production Open Source RAG Application using LlamaIndex and Cassandra
01h 59mLLM applications with LangChain
17mArchitecting LLM-Based Chatbots using free open-source LLMs
27mAI Demystified Unveiling the Intelligence Behind the Machine
35mUnleashing Creativity with Spring AI
47mRamaLama: Making working with AI Models Boring
01h 34mHow to Construct and Assess Your First LLM Based RAG System