Security-focused software engineer. I build verification-first tooling — systems where an AI or a user can propose, but deterministic code, policy, and cryptography decide — and I do independent, authorized bug-bounty research on the side.
Every project below carries a ## Status section stating what is implemented, tested, live-verified, experimental, or historical. Nothing here claims production readiness unless demonstrated.
- sealbox — TypeScript CLI that seals a file to Shelby storage with a canonical, Ed25519-signed manifest and verifies it independently. v0.3.0; 174 tests, CI green. Upload path live-verified; live retrieval blocked on an upstream anonymous-tier limit (stated honestly as
PARTIAL). - nexescrow — Non-custodial milestone escrow on EVM (Solidity 0.8.26, Foundry, Next.js). 57 Foundry tests incl. reentrancy PoC and fuzz invariants; testnet-only, unaudited; Demo 1 hardening (EIP-712 intents, resolver modules) is an open, unmerged PR stack. BOT Chain testnet evidence is kept separate from network assumptions.
- Morphia (private, hosted alpha v0.2.1) — orchestration for authorized security engagements: code-owned tool registry, live scope enforcement, human approval gates, signed execution receipts, evidence provenance. Not production-hardened for multi-tenant use.
- postalmind-ai — bilingual (Tamil/English) public-interest evidence tooling; deployed and CI-smoke-tested.
- quantum-foundations-lab — executable lab rebuilding 3-qubit teleportation from the postulates so maths, code, tests and interpretation cross-check (v0.1.0, CI green).
- proof-of-reasoning — historical, failed hypothesis: a 2026 design for a consensus-scored reasoning subnet, preserved with a criticism matrix and the lessons that moved later work toward deterministic verification.
- Protocol-architecture research corpus (Bitcoin, Ethereum, Monero, Babylon, Nexus and others) is maintained privately; nothing is published from it yet.
Hackathon-week reference implementations, each with CI and a stated scope ceiling: actionharbor (policy-gated action ledger, 379 tests) · controldeck (governance-vs-orchestration demo) · sourcelens (evidence-grounded RAG) · shiftpilot-ai (NL→deterministic scheduling, hosted) · pi-runner (C++→WASM game, deployed).
safecheck-pharos (hackathon submission, 2026-07) · defi-vibecoding-data-analysis (Arbitrum Trailblazer dataset) · online-food-ordering-cpp (academic). Archived placeholder repositories are marked as such and are candidates for deletion.
TypeScript / Node.js · Python / FastAPI · Solidity / Foundry · Next.js / React · C++ → WebAssembly · PostgreSQL · Docker · GitHub Actions.
Rishidar-lab is the owner, originator, and sole human creator of these projects unless a repository states otherwise. AI assistants (including Anthropic Claude) are used as development tools for research, coding, review, testing, and documentation; direction, acceptance of output, and responsibility remain with Rishidar-lab. Third-party code and data retain their own licenses.
Bugcrowd · X / Twitter · Email