Enterprise AI, data platforms, and the side projects that test my ideas
SentinelAI: a governance gateway for enterprise LLMs
A gateway between employees and LLMs that blocks prompt injection and sensitive-data requests, and logs every decision for audit. Measured on a 200-prompt eval set.
Sector: a weekly planning app organized by time
A phone-first task app where time is the only way to organize: five buckets, a weekly limit and a short weekly review. Built from discovery and a PRD to a live app.
Tierwise: EU AI Act risk tiers inside Jira
A Jira app that gives each AI system an EU AI Act risk tier from 6 questions, and blocks high-risk work until evals pass and a reviewer signs off.
CaseCraft: a practice room for PM interviews
Pick a case, face an interviewer, beat the clock, get scored. 87 cases across 9 PM interview types, live as a front-end demo.
Aura: from crowded to calm
A PM audit and redesign of Aura, an AI study tutor with spaced-repetition flashcards: honest due counts, a recall-first study flow and a proposed Today screen.
From interview to rollout.
I work closely with business teams, engineers and leadership, turning interviews and usage data into specs a dev team can build. On my own projects I run the same path end to end, from discovery and PRD to a shipped app.
Listen before building
I talk to the people who use and fund the product, from engineers to executives, before deciding what to build and what to cut.
Let usage data decide
I prioritize by how people actually use the product, so the roadmap rests on evidence, not opinions.
Prototype early, learn fast
I get a rough working version in front of real people quickly. Their reactions settle debates faster than meetings do, so we iterate or move on before over-investing.
Build the AI myself
I stay hands-on with AI agents, RAG, LLM evals and guardrails, from personal projects to enterprise solutions.