At Xpression, I’m building an AI app builder that turns ideas into working full-stack apps. I solo-built its context service, replacing vector RAG with relationship-aware retrieval that reduced token use and cost per call while improving judged context accuracy.
I also solo-built a DAG-based, six-agent runtime with parallel execution, retries, and checkpoint resume. For long sessions, I added priority-based memory compaction; one representative session went from 95K to 63.4K tokens while keeping auth and compliance rules verbatim.
At JUTEQ, I built and deployed an AutoTrader Canada customer-service chatbot used by 10K+ users, and contributed to voice agents and browser automation for dealership scheduling. Earlier, as a Back End Developer intern at Indian Institute of Technology, Mandi, I built a sensor-data service, REST APIs, and a live monitoring dashboard.

