Ripon Chandra Malo's profile on Product Hunt

About

I'm Ripon, creator of Matily — a small, independent open-source studio. My core work is projectmem: a local-first memory and judgment layer for AI coding agents. It remembers your debugging history, failed attempts, and decisions, then warns you before you repeat a mistake. I also build GemType, a free, open-source Grammarly alternative. Everything I make is free, private, and runs on your own machine — no accounts, no tracking. Good software should be free, private, and open to all.

Links

Matily

Twitter

Badges

Tastemaker

Tastemaker 5

Gone streaking

Gone streaking 5

Maker History

Jharu -A disk cleaner that can think.Free disk cleaner for your AI model and dev caches
Jul 2026
GemTypeFree, open-source Grammarly alternative powered by Gemini
Jul 2026
projectmemMemory + judgment for AI coding agents (local, MIT)
May 2026

🎉

Joined Product HuntMay 18th, 2026

Forums

•

29d ago

Jharu -A disk cleaner that can think. - Free disk cleaner for your AI model and dev caches

Jharu is a free, open-source disk cleaner for macOS and Windows that understands AI and developer clutter. Reclaim gigabytes from Hugging Face, Ollama, and PyTorch model caches, npm and pip caches, and app leftovers. A free CleanMyMac alternative — no telemetry, nothing permanently deleted.

•

1mo ago

GemType - Free, open-source Grammarly alternative powered by Gemini

GemType checks your grammar and rewrites text on any website — Gmail, LinkedIn, X, Reddit, anywhere. Like Grammarly, but free and open source: you plug in your own free Google Gemini API key. No account, no subscription, no tracking.

•

2mo ago

Hi Product Hunt 👋 Solo dev + grad student building memory for AI coding agents

Hey everyone

I'm Ripon a grad-student researcher at the University of Utah and the solo developer behind projectmem.

What pulled me into this: my AI coding agent kept forgetting everything between sessions. Every morning it re-read the same files and cheerfully re-suggested a fix I'd already tried last week one that had already failed. The model wasn't the problem. The missing memory was.

So I built projectmem: a local-first memory + judgment layer that lets an AI agent remember what was tried, what failed, and what was decided and warn you before you repeat a known dead-end. 100% local, MIT, no telemetry, no cloud. It's on PyPI and the MCP registry now, and there's even a research paper behind it.