--- title: "About — Samrat Mukherjee" description: "Who Samrat Mukherjee is, how he works, and why reliability engineering shapes the way he builds LLM agents." url: "https://0xsamrat.com/about" author: "Samrat Mukherjee" updated: "2026-09-18" --- # About Samrat Mukherjee > AI Engineer · AI Agent Developer. I build LLM agents that take real actions for users. AI engineer with 4 years building production systems, now focused on LLM agents that take real actions for users. Built and shipped two live AI products, one with paying customers, where agents research, decide, and act autonomously. Strong on the parts that break in production: RAG grounded in user data, anti-hallucination guardrails, LLM-as-judge evals, and full tracing with LangSmith and OpenTelemetry. The reliability discipline comes from years on financial infrastructure in Go — settlement systems where idempotency, crash-safety, and never double-charging are non-negotiable. That discipline now applies to agents that cannot take an action back. ## How he works The through-line across every system here is the same: an agent is only as useful as the worst thing it does unsupervised. So the interesting engineering is rarely the prompt — it is the boundary around the model. - **The model decides, the code computes.** In Stockpot the matching engine is fully deterministic and agent tools accept no model-supplied numbers: the model chooses which check to run, Python computes every value. - **No fact, no answer.** Shortlistapp's FactGuard layer blocks the model on any field with no retrieved fact, so the agent answers from real data or escalates to the user. It never fabricates on a submission that cannot be taken back. - **Untrusted data is data, never instructions.** Icebreakr isolates scraped profile text in delimiters with a typed-error retry loop on failed checks. - **Measure the agent, not the vibes.** LLM-as-judge scoring and prompt-regression tests run before merge; LangSmith and OpenTelemetry traces make a bad run reproducible instead of anecdotal. ## Where the discipline comes from Four years of backend work before the agents, most of it on money movement. At Atum Labs it was stablecoin clearing and settlement: an event-sourced Postgres store with a race-free two-phase work queue, crash-safe and idempotent across a 20-worker pool. At Defipe it was a derivatives execution engine — order matching, margin accounting and liquidation, validated across 10K+ simulated leveraged trades with zero false liquidations. Settlement systems taught the same lesson agents need: the retry is the dangerous path, and "probably didn't double-charge" is not an answer. That is the habit now applied to agents that take actions users cannot undo. ## Timeline ### 2022 - **Feb 2022 · Joined** — Started at TechExactly. First gig in tech. Smart contracts, a React frontend, and the habit of writing tests for the failure paths. ### 2023 - **Feb 2023 · Winner** — Won ETH for ALL hackathon. First hackathon win — with the project that would become Defipe. (https://devfolio.co/projects/defipe-b5ba) - **Sep 2023 · Founded** — Founded Defipe. A derivatives trading protocol in Go. Off-chain execution engine, on-chain settlement. - **Sep 2023 · Winner** — Won ETHGlobal — Connext Pool Prize. Ape DAO 2.0: cross-chain governance with multichain proposal and voting execution. (https://ethglobal.com/showcase/ape-dao-2-0-i846k) ### 2024 - **2024 · Grant** — Soonami Venturethon, Cohort 5. A $1,000 grant for Defipe, plus three months inside Soonami's network of founders and builders. (https://app.foundance.org/projects/11247) - **2024 · Milestone** — Defipe → 500+ monthly active users. Peak usage, and a 1,000+ member Discord around it. ### 2025 - **May 2025 · Wound down** — Shut down Defipe. End of a two-year run. Took the lessons on idempotency and crash-safety with me. - **Aug 2025 · Joined** — Backend engineer at Atum Labs. Go services for stablecoin clearing and settlement — Kafka auctions, gRPC streaming, event-sourced Postgres. ### 2026 - **Jan 2026 · Founded** — Started Shortlistapp.co. An autonomous job-application agent. First paying customers, and the first agent I shipped that takes actions a user cannot undo. (https://shortlistapp.co) - **2026 · Launched** — Icebreakr on the Chrome Web Store. AI outreach with multi-provider fallbacks, prompt-injection hardening and an LLM-as-judge eval harness. (https://useicebreakr.com) - **2026 · Shipped** — Stockpot · Agentic Commerce Hackathon. Agents trading surplus food between restaurants, settling on real payment rails. 749 tests, including an adversarial suite. (https://getstockpot.shop) - **2026 · Launched** — NetagiriFiles. Civic transparency for ~800 Lok Sabha MPs, serving 500+ monthly visitors. (https://netagirifiles.fun) ## Education B.Tech, Electrical Engineering — B. P. Poddar Institute of Management and Technology, 2017 – 2021 ## Hackathons & achievements - **Soonami Venturethon (Cohort 5)** — Won a $1,000 grant for Defipe.io; grew it to 500+ testnet users and a 1,000+ member Discord. (https://app.foundance.org/projects/11247) - **ETHGlobal — Connext Pool Prize winner** — Cross-chain DAO governance with multichain proposal and voting execution. (https://ethglobal.com/showcase/ape-dao-2-0-i846k) - **ETH for ALL — Arcana Pool Prize winner** — Decentralized token swap via 0x Aggregator with social-login wallet onboarding. (https://devfolio.co/projects/defipe-b5ba) ## Contact - Email: samrat.mukherjee2022@gmail.com - Book a 15-minute call: https://cal.com/0xsamrat/15min - Location: Bengaluru, India (Asia/Kolkata, UTC+5:30) - Availability: Open to work — full-time, remote-friendly - Resume (PDF): https://drive.google.com/file/d/14feSsx0a-vZ4A2sf8XZluPPLiVPx2DK_/view Reply time is same day, usually within a few hours. There is no contact form — email or the booking link are the two working paths.