Himadri Mishra

Himadri Mishra

Agent Experience Engineer. AI Product Engineer.

I build tools agents can find and use. I ship the access, evaluation and reporting systems behind them.

Explore my work Email HimadriDownload resumeGitHub

Selected engineering

Agent systems I have built.

Shipped evaluation platform · completed studies

Agent journeys, encoded as state.

I turned onboarding journeys into executable capability states, then used them to compare tool routing and task completion across agent clients. I built the evaluation infrastructure, state-matched router cards and evidence capture needed to inspect which journeys worked better.

Capability-state contracts, multi-client harnesses, matched A/B tasks and trace-backed outcome checks.

Explore the evaluation engineering

Anonymized engineering summary; underlying work is private.

Inside the evaluation
  1. Encode the starting stateInstallation, access and available tools
  2. Compare matched journeysControl and treatment, across clients
  3. Inspect the outcomeTool choice, task completion and failures

Simplified method illustration, not a production trace.

Shipped system

MCP access and onboarding.

I shipped hosted MCP authentication across account-connected, search-only and keyless entry points, connecting authorization, token lifecycle and CLI setup with explicit recovery paths.

OAuth, scoped access, headless setup, token and grant lifecycle, actionable recovery.

Explore access and recovery

Anonymized engineering summary; underlying work is private.

Shipped measurement and reporting

From discovery to grounded insights.

I built measurement for whether agents could find, retrieve and use a product, then built evidence-grounded insight reporting on top of those observations. The work connects evaluation banks and retrieval probes to synthesis, validation and recoverable reporting.

Category, developer and goal-led evaluations, retrieval APIs, evidence bindings and checkpoint recovery.

Explore measurement and insights

Anonymized engineering summary; underlying work is private.

Shipped working versions

More shipped agent systems.

Explore the implementation and engineering decisions behind each system.

Anonymized implementation summary; underlying code is private.

The complete engineering portfolio

Full technical detail, scoped results and broader work across backend, frontend, ML and computer vision.

View all engineering work

Writing

Agent Experience field guide

Practical guides to discovery, tool use, evaluation and recovery, with a read-only MCP service for searching and retrieving the content.

Independent guidance, with explicit human-control boundaries.

More writing and open-source work

About

My earlier work spans automated research, ML infrastructure, search and computer vision. I bring the same questions to agentic systems: does it work, what does it cost, how does it fail, and can the next engineer operate it?

Contact

Send me what you are building.

himadri.jobhunt@gmail.comGitHubResume PDFX / TwitterLinkedIn