"The future belongs to those who can master the art of conversation with machines." Ajay Tiwari
|
Structure aware chunking, embeddings matched to the domain, and a wide vector search narrowed by a reranker. Follow up questions are rewritten with conversation memory, and answers stay grounded or abstain instead of guessing. |
Stateful, tool using agents orchestrated as graphs and connected to real systems through MCP servers, APIs and databases. Served with FastAPI behind signed webhooks, with memory, retries and background work, on whatever channel the users already live in. |
Every stage of a RAG pipeline or agent is scored in isolation with DeepEval, then end to end with LLM as a judge rubrics. Regressions fail the release. In production, LangSmith scores live traces continuously and alerts on any metric drop. |
| ๐ง Agents & Frameworks | |
| ๐ฎ Models & Providers | |
| ๐ RAG & Vector Stores | |
| ๐ก๏ธ Eval, Guardrails & Observability | |
| โ๏ธ Backend & APIs | |
| ๐๏ธ Data | |
| โ๏ธ Cloud & DevOps |


