Harshit Sharma
AI Product Manager
Bridging engineering depth with product vision. Building the next generation of AI-powered tools at the frontier.
Background
Building AI Products at the Frontier
I'm transitioning into AI Product Management after 5 years at GE Vernova, where I built complex products with engineering teams and shipped at scale.
Currently completing my MEng at UC Berkeley (IEOR), I'm targeting PM roles at frontier AI companies like OpenAI, Anthropic, and leading AI startups.
I bring technical depth, strategic thinking, and a hunger to solve the most interesting problems in AI.
UC Berkeley MEng
Industrial Engineering & Operations Research
5 Years at GE Vernova
Complex product development at scale
Frontier AI Focus
Targeting OpenAI, Anthropic & leading AI startups
Selected Work
Explainable AI Coding Assistant
An AI-powered code completion tool featuring calibrated confidence scores and explainable reasoning. Tackles the critical “trust tax” in AI coding assistants by showing developers why a suggestion was made and how confident the model is, reducing blind acceptance and manual verification overhead.
The $3B+ AI coding assistant market faces a core adoption barrier: developers don't trust AI-generated code. This prototype solves that with a confidence-calibrated system that surfaces reasoning inline, letting developers make informed accept/reject decisions at a glance.
Calibrated Confidence
Reduces blind trust and speeds adoption
Codebase Intelligence
Beyond naive RAG for repo-specific reasoning
Legible Correctness
Visible during coding, not just in post-review
Perspectives
Hot Takes & Random Facts
Most AI products fail not because of bad models, but because of bad product thinking.
The best PM skill is saying no to features that demo well but ship poorly.
RAG is necessary but not sufficient. Context without reasoning is just expensive search.
Connect
Let's Build Something
Interested in discussing AI product opportunities, collaborating on research, or just want to talk shop.