Indie developer · AI security · shipped to real users
MCA grad from Assam, India — building solo across AI agents, security tooling, dev tools and mobile apps.
/\_/\
( •ᴗ• ) ▸ status: SHIPPING
> ^ < ▸ stack: rust · python · typescript · flutter
▸ base: assam, india
▸ stacktrace: 99% cat naps, 1% shipping
Most of my projects start the same way: "Wait... why doesn't this already exist?" — then I lose a few evenings building it.
|
Zero-trust execution layer for AI agents — every tool call is authorized against an explicit default-deny policy and recorded in a tamper-evident audit chain. |
Published CLI that audits AI-generated repos for quality and security risk — deterministic checks, SARIF output, fully local-first.
npm · source |
|
Free exam-prep app on Google Play for Army / Assam Police / SSC / CAPF aspirants — full mock tests with real negative marking and in-test English / Hindi / Assamese translation. site repo |
Clean USB Rescue Engine — finds how software persists across reboots (Run keys, services, WMI…), attaches evidence to every finding, quarantines but never deletes without consent. |
|
Research harness behind my first paper (in prep) — measures Kavach against 42 real prompt-injection attacks from IssueTrojanBench, static analysis plus live baseline-vs-defended agent runs. |
Local-first environment for evaluating AI agents — grid-world planning plus protocol adherence inside a deny-by-default capability sandbox. It ships the audit that says how far that currently goes. |
Also: WireSong — network traffic as generative soundscape (live demo) · Pehredar — Android root/spyware detection over ADB · Ustad — local LLM distillation studio · → all repos
mail me if you're building agent-security or devtool stuff.
/\_/\ ( ^ᴗ^ ) > ♥ <
thanks for scrolling this far. cat approves. star button bhi upar hi hai ↑ — just saying.
probably coding something that wasn't on the roadmap.
still learning · still building · still shipping





