About

I’ve spent a decade building AI systems for environments where the demos that excite people on LinkedIn wouldn’t survive a single week in production.

In 2019, sitting at Toyota Motor Europe in Brussels, I shipped a production NLP system that turned multilingual car owner comments into something the engineering teams could actually act on. After engineering came strategic and implementation consulting, where I led LLM extraction pipelines, ticket intelligence systems, and predictive quality models for global automotive OEMs. Most recently the craft has been applied to financial due diligence — a workflow that still consumes weeks of analyst time and hasn’t meaningfully evolved in two decades.

The through-line is regulated environments: industries where data is sensitive, mistakes have consequences, and “we’ll just use frontier LLM models” is not a viable answer. Automotive, industrial, financial — different domains, same deployment problem.

Work and education have taken me through Dar es Salaam, Dubai, Melbourne, Aachen, Brussels, and Munich, usually in whichever language the room is working in. I’m based in Munich now, working across Europe and the Gulf.

The lesson from all of it: real-world AI fails not because the models are bad, but because no one has bothered to understand the workflow it’s replacing. That’s the gap I build into.

Reach me at hi@pulkitgaindhar.com.