About
I lead AI and ML at TravClan, India's B2B travel tech platform, where I build and ship the production systems behind pricing, quoting, and customer operations. That means Atlas, a quote system that turns raw agent requests into bookable itineraries; a dynamic pricing engine running across hundreds of partner agencies; agentic systems handling flight operations and support tickets; voice AI in live sales and support calls; and the data infrastructure the rest of the company depends on. Before this I built TravClan's analytics function from the ground up. Outside work I write The International Strategist on Substack.
How I work
The systems I build sit in front of money changing hands, so most of the engineering is in deciding what a language model should not be trusted with. In Atlas the model is load-bearing exactly once — turning an agent's prose into a structured request — and everything after that is a solver, a live inventory lookup, or a validation stage that is allowed to refuse. Getting that boundary right matters more than any prompt.
The corollary is that evaluation is the real work. A system that fails loudly is a bug; one that fails quietly is an outage you hear about from a customer. Most of what I have learned in production came from building eval sets out of real inputs rather than synthetic ones, and from metrics that measure coverage rather than error rate — because the expensive failures return a success code.
Before this I built TravClan's analytics function from the ground up. That era was automation and classical modelling rather than language models: automation work across operations that cut manual effort by around 30%, and fraud detection models that reduced fraud risk by around 36%. It is why I reach for a classical model when one will do, and why I treat an LLM as one component in a system rather than the system itself.
Background
- AI and ML across pricing, quoting, and customer operations
- Analytics for Product; built the dynamic pricing analysis framework
- Automation, BI, and fraud detection modelling
Recognised at TravClan across five consecutive years — Pillar of Excellence (2025), HIRO Award (×3), HIRA Award (×4), CX Star
B.Tech, Electrical and Electronics Engineering · SRM Institute of Science and Technology