Luvish Gulati
Product-minded software engineer & AI builder
Hey — I'm Luvish. I own products end to end: discovery and PRDs through code, launch, GTM and analytics. Right now I'm a Software Engineering Intern at GrowthX — I brought community notifications to mobile end to end (unified in-app + push gating, per-channel preferences, native banners), shipped an AI reviewer that gates incomplete event listings, ran the app-install GTM across transactional emails, and built a device-level E2E harness that asserts push-vs-silence against the OS itself.
Before that I was an Associate Product Manager at Pink Unicorn Algorithms, running three products — CarbonNex, Rapidero and Risk AI Council — and an Associate Product Consultant Intern at Cvent across 10+ enterprise accounts.
My main build is Henry — an open-source, local-first AI chief-of-staff with a human-like memory engine (semantic + lexical recall over an associative graph, consolidated nightly), cited RAG retrieval over my GTM and project-management playbooks, a Telegram standup bot for my team, and a morning job scout — on an 8 GB M1 Air at ~$0 marginal cost. Nothing goes outbound without my sign-off, and it maintains and deploys this website itself.
work experience.
featured projects.
-
Henry Flagship
Local-first AI chief-of-staff that remembers, schedules and drafts — and never sends without sign-off.
-
AI Digital Twin for Recruiters
Context-aware chatbot recruiters can interrogate about my work instead of reading a static resume.
-
Coding Efficiency Measurement System
Grades the coding process, not just the output — a Net Efficiency Score from keystrokes, idle time and focus intervals.
case study · private repo
-
Risk AI Council
AI-risk platform where I led GTM and content across 6 risk categories, aligned to NIST AI RMF and the EU AI Act.
-
CarbonNex
ESG reporting product — I designed the core workflows as its associate product manager.
technical skills.
Engineering
AI
Product
github.
Luvishgulati03 ↗166 contributions in the last year
LessMore
henry stats.
ai-agent- ↗62
Queries answered — one memory logged per turn
140
Memories formed — episodic, semantic and procedural
<1 month
Active for — running as a daily driver
$0
Marginal cost per run — no API keys · runs on an 8GB M1 Air
as of 8 Aug 2026 · read straight out of Henry's local databases, never hand-typed