At Neuronest AI Pvt Ltd, I developed real-time conversational voice agents and connected RAG tools to ground their responses in unstructured documents.
I also built asynchronous FastAPI services for LLM and tool workflows, with reported API response times under 200 ms.
At Admire Solution, I developed transformer-based and agentic AI workflows that improved downstream task accuracy by 28%. I optimized inference with quantization, dynamic batching, and edge-deployment techniques.
In my projects, I orchestrated support and product-discovery agents for RetailPilot and published ScopeForge, a Python CLI/TUI research framework with a RAG knowledge base, persistent memory, and golden evaluations.

