Backend engineer building distributed systems and AI developer tooling.
- Trust Identity Protocol (TIP) at theailaborg: a content-provenance platform with cryptographic signing, juror-based dispute resolution, and Merkle-anchored canonical state. I work across the whole ecosystem: the public protocol core, plus the SDK, browser extension, AI classifier, protocol node, mobile web app, and WordPress plugin in private org repos. That work happens on my work account, @t-bhendarkar, which is where my daily contribution graph lives.
- ML and data tooling at the same org: PII redaction, an LLM inference toolkit, research-paper extraction and metadata pipelines, and audio transcription (private repos).
- Ship & Tell: an MCP server and Claude Code skill that mines your local AI coding sessions and turns them into tweets, threads, and articles. No external API: the coding agent's own model does the writing, and nothing leaves your machine.
- Async classifier pipeline: moved a slow classifier out of the request path so the API returns immediately and the two halves fail independently
- Canonical-state reset fix: made clear and build functions exact inverses so Merkle roots stay consistent across nodes
- Subagent transcript mining: found that Claude Code stores 5x more transcripts than the obvious glob shows, and built dedup-aware reading on top
Python, TypeScript, Node.js, React, Next.js, Postgres, Redis, Laravel.
- X: @tusshhar_r
- LinkedIn: tushar-bhendarkar
- Website: tusharbhendarkar.com
- Email: tusharbhendarkar44@gmail.com


