Hey, I'm Tirth 👋
AI engineer in India, founder, and team builder.
Close to a decade across the stack: frontend, backend, infra, design, and now AI systems end to end. I work from India with teams building model-powered products that need to make it past demo-quality.
For the last five years, AI systems have been the bulk of my work, anchored at BigCircle, the studio I founded. That work is mostly production LLM systems: agents, MCP tools, retrieval pipelines, evals, fallbacks, and the full-stack infrastructure around them.
I've taken agentic systems from prototype to production, both on products of my own and on partner engagements. The niche is not "AI feature" work in the abstract. It's context engineering, tool design, evals, and cost-and-latency-aware systems that hold up under real traffic.
Writing and thinking these days about agent architectures, eval-driven development, and the distance between an LLM demo and a system that stays up under load.
What I ship
- Production agent systems and LLM features that need to survive real traffic.
- MCP tools, retrieval pipelines, evals, and fallbacks around model calls.
- Full-stack product work when the AI is the hard part, not an ornament.
Signal check
- Base
- India, working globally.
- Studio
- Founder at BigCircle.
- Recent systems
- Nurix, UBIK, Brandiligence, plus earlier work at Gartner and Blueland.