Pulkit Verma
AI Engineer, frontend-minded systems builder
Selected portfolio / 2026
I work on agentic workflows, productized AI systems, and polished interfaces that make complex software feel inevitable.
Principles
Determinism > randomness
Agentic SDLC, not AI theater
Zero-hallucination is a systems problem
What you can pull me in for
Workflow design, orchestration, eval loops, and shipping paths.
Interfaces that reduce cognitive load while still feeling sharp.
Review automation, internal tooling, and systems that scale teams.
Featured builds
The best entry point is a mix of shipped product, internal leverage, and infrastructure thinking. These cards should make that legible in under a minute.
Production-oriented AI agents and RAG systems designed for teams that need reliability, not demos.
Visual feedback capture with browser context, built to reduce the drag between finding issues and shipping fixes.
Autonomous code review against team principles, pushing AI into the quality loop of the SDLC.
Retrieval-first architecture work aimed at reducing hallucination through system design, not prompt tricks.
“The interesting work isn’t making AI talk. It’s making AI behave.”
Technical notes
This should feel like a working notebook, not a generic blog. Fast to scan, opinionated, and worth bookmarking.
A piece on why most agent failures are orchestration failures: weak state boundaries, vague retries, and invisible handoffs.
Notes on using editorial hierarchy and restraint in technical interfaces.
Read noteRetrieval quality, chunking, and evaluation loops matter more than prompt folklore.
Read noteBrowse archive
A full notes index can become the long-tail trust layer of the site.
View all notesResume snapshot
The site can frame you as a systems builder first, while the resume can sharpen the frontend engineering story for hiring and outbound.