AI and product engineer. I build and ship, fast and daily: full-stack products, AI agents, automation pipelines, and the tools around them.
Open source: Google Summer of Code, Linux Foundation (LFX), and Symmetry Autumn of Code. 3+ years shipping in production.
Products and apps
- SignalizeAI: AI prospecting as a Chrome extension, featured on the Chrome Web Store.
- AashiNyra: an AI fashion-influencer platform I run end to end, built on an n8n automation layer that drives the growth loop.
- Kabootar: Offline mesh messenger, chat with no internet, no servers, no SIM. Messages hop phone-to-phone over Bluetooth and Wi-Fi (delay-tolerant, epidemic routing, end-to-end acks). Flutter.
- Tiny Day: A cozy, offline-first daily planner for Android, no accounts.
- PennyRush: A private money hub, receipts and statements parsed in memory, only extracted fields stored.
- Auralis and Mycelia: GPU experiments in the browser, a real-time WebGPU aurora and a 262k-agent slime-mold simulation, no server.
Open-source lists and skills
- Curated, auto-updating directories I maintain: awesome-agent-evals, awesome-llm-guardrails, awesome-mcp-security, awesome-context-engineering, awesome-agent-memory, awesome-agent-skills, plus the distro-skills Agent Skills pack.
AI agent infrastructure
AgentPostmortem is the flagship: a public registry of real AI-agent failures. Around it sits a suite built on one idea, that with agents generation is easy and verification is the hard half:
- Evals and observability: Agentrace, Evalgate, Tracecase, Voiceeval
- Retrieval and trust: VaultRAG, Tenantq, Answerproof
- Security scanners: MCP-audit, Skill-audit, Injection-arena
- Context and token cost: Ctxlens, Ctxtrim, tokencut
- Scoped MCP access: Bridgekit, Casebook MCP, Casebook Chat
- Human-in-the-loop agents: Resolvd, Greenlite, Webhands, RelayG
Roots
- cpp: Where I started, learning C++, datatypes through graphs and DP.