I'm Samir Maikap. I help founders and product teams turn an AI idea into a product people rely on, in weeks. You get one senior engineer who plans with you every week, builds with AI coding agents, and proves it works with tests and evals before it ships.
AI products I've built and shipped. 49 projects across 175 repos since 2017. Code walkthrough on request.
Try-on clips rendered on the shopper's own selfie, with a one-shot fallback when a model refuses. The catalog grows from what shoppers search for.
Lean tools in 2018, then LeadershipFITT with an AI coach grounded in each user's results, now FITTWorks.
Capture decisions from web, email and Slack, frame and forecast them, follow up on outcomes and review.
Companion app for the Bartesian cocktail machine: capsules, recipes, shop, orders, rewards and subscriptions, across web and IoT.
Field app for logging wildlife sightings with photos, GPS location and maps that work offline.
Slack-native team checklists running on Claude Haiku.
Mac menu bar app that cleans up what AI coding agents leave running.
No duplicate posts on retries, an MCP server for AI agents, tested tenant isolation.
AI assistant that answers from a company's Google Drive and cites its sources.
Also built for clients: Waywego (travel, Gemini), Vitrine, UpkeepSkin (Shopify), Africa One Stop, Slick-Rx (led 5 devs), The Biggest Ask, Rumo, CarBye, Korean B2B and Expensense.
I'm a Senior AI Product Engineer based in Kolkata, India, working US hours. Since 2017 I've built products end to end for US founders and agencies: 49 projects across web, mobile and AI, including the Bartesian cocktail-machine app, Catalog's AI stylist, and the FITT product line I've led for MCS Media for eight years.
Teams hire me when an AI feature has to work in production, not just in a demo. I scope the work in writing, build with AI coding agents, and measure AI output with evals before it ships. On Decize, that process took faithful debrief lines from 91.6% to 100% on a fresh holdout.
With the founders: Catalog (AI fashion), Feather (AI knowledge base) and The Quest (iOS game).
Lean Management Systems. Three generations of the same product line for one client.
B.Tech, Computer ScienceWest Bengal University of Technology · 2016
What I ship with, and the projects where I used it.
A debrief sums up what happened after a decision. If the model invents a single line, nobody trusts the ledger again. This is how I measured that risk and fixed it.
Scores count the debrief lines the model showed. Read the full write-up →
Team-sized output from one accountable engineer. You bring four things; you get four back.
Your ideaOne paragraph is enough to get started.
A deadlineA real date lets me plan the scope around it.
A budget rangeA rough range is fine. I'll tell you what fits inside it.
Weekly planningOne call a week with you. I own the rest, all the way to production.
A live productWeb, iOS or Android, live on your domain or in the App Store.
All the source codeIn your GitHub account from the very first commit.
Automated testsLint, type checks and tests run in CI on every change.
AI evalsAI output is scored against a bar set before the run.
From idea to a live web or mobile product.
RAG search, agents, MCP servers and vision, added to what you have.
One engineer who owns your product line for years.
Previously we had worked unsuccessfully with 8 other app developers… Samir took the initiative with many of the UX issues… [and] was available any time we needed him.
Send your idea, your deadline and a rough budget.
maikap.samir@gmail.comAvailable now for full-time remote roles, with 9am–1pm ET overlap. Code walkthrough on request.