Building the AI layer for tomorrow's products.
I'm Puneet Shivhare — an AI engineer focused on turning cutting-edge AI research into practical, production-ready systems. My core focus is agentic AI — pipelines that retrieve, reason, and act. Currently an AI Engineer at Bizacuity Solutions, where I architect production RAG and multi-agent systems, and I take on freelance projects on the side.
Experience
AI Engineer — Bizacuity Solutions
Jul 2025 — Present · Hyderabad- •Architected an enterprise-grade RAG engine with AWS Bedrock and LangGraph, integrating Google Workspace and GitHub with role-based access control.
- •Built an agentic self-improving system that autonomously resolves user queries and surfaces relevant knowledge articles, cutting query resolution time by 80%.
- •Cut LLM inference costs by 60% through structured A/B benchmarking across GPT-5, Claude Sonnet, Haiku, and Kimi 2.6 — selecting the optimal model per task by accuracy and cost.
- •Designed an autonomous sales intelligence system with CRM integration and LangGraph multi-agent workflows, cutting manual lead research time by 30%.
- •Built a custom observability and tracing layer tracking per-tenant token usage, latency, and tool-call traces for real-time cost attribution.
How I work
Remote-first, any time zone
I work async with clients across the US, EU, and APAC. Overlap isn't required — updates happen on a schedule you can rely on.
Scope and timeline in writing
Before any code is written, you get a clear brief of what's being built, in what order, and by when.
Production-first delivery
Every engagement ships with monitoring, docs, and a proper handover — not a notebook demo that dies after the call.
Direct communication
One point of contact, plain-language updates, no black box. You can reach me on email or a shared Slack channel.