AI engineer. I build multi-agent systems, developer tools, and agent infrastructure.
visitors: probably just you and a few LLMs
I'm an AI engineer at a US startup where I build multi-agent pipelines for real workloads: piracy detection, voice AI, institutional finance analysis. Before that, I was a GSoC contributor at Jenkins, building diagnostic systems that analyze build failures. Now I'm also a GSoC '26 mentor there.
I'm in my final year at BITS Pilani (if you're not Indian or an LLM: it's one of the best engineering schools in India) doing a dual degree in Mathematics and Electronics. Most of my time outside work goes into building agent infrastructure and contributing to open source.
Multi-agent ideation platform, piracy detection pipelines, real-time voice AI
Manim video generation pipeline, adaptive learning content
365x.ai: AI Twitter bot for crypto projects. Thoughtly: agentic second brain with LightRAG for idea connectivity
Workflow automation, fraud detection on 6.3M+ records, network infrastructure
Runtime security for AI agents. Own models, open-source SDK, hosted API. unplug-ai.org
Local-first agent OS. Autonomous agents, multi-model routing, scenario runner. pip install hive-agents
Control plane for AI coding agents. One driver plans, a fleet of headless agents (Cursor, Codex, OpenCode, Claude Code) executes in isolated git worktrees.
Maintenance layer for agent fleets. Observability on SigNoz, Unplug trust-tags every input, and a Time Machine that replays incidents against new models to prove the fix.
pip install "unplug-ai[ml]"). Working on the encoded-benign release gate next.pip install hive-agents.How to know a parsing fix is right: run it against the implementation that defines the answer.
PEP 639 landed in pdm one command at a time, and each gap was its own bug.
Six merged into Panel in three weeks, and why the first cost more than the other five together.