AI Product Engineer İzmir · open to İstanbul
Product-minded AI engineer. I pick up a new domain fast, turn it into a working product quickly, and stay accountable for it after launch. So far that has meant physiotherapy motion analysis at Propose, industrial process safety, structural damage vision and an app live on the App Store. I work best inside a team, owning my part end to end.
Ask me about , , or .
About
I'm Tarık, an AI engineer who thinks in products. I start from the user's problem, not the model, and carry the work through backend, interface and deploy until it is live and in use. The domains keep changing: physiotherapy video at Propose, HAZOP risk studies, bridge damage segmentation, nutrition and training. The way I work stays the same: learn the domain quickly, ship a first version early, and take responsibility for what I ship. My time on Hepsijet's backend team showed me how much further that goes with code review, CI/CD and a team around you, and that is where I do my best work. I also bring real depth in sustainability and climate tech, from carbon-accounting platforms to a formal ESG standards program (ISO 14064, GHG Scope 3). Based in İzmir, open to İstanbul and remote.
Where I studied, and the roles that each left a measurable mark.
Part-time, remote, at a physiotherapy motion-analysis startup. I build an automated pose-estimation pipeline that turns public exercise video into a 1,000-exercise clinical library, a Bedrock vision auditor that auto-approves clean clips and sends only flagged ones to physios, and grantos, a grant-intelligence engine that screens 400+ funding programmes daily. I also adapted my own outreachos into a prospecting tool that finds UK and EU physiotherapy clinics and their decision-makers (built, not yet in active use).
est. 9 → 2.5 weeks per library · 180 tests · 400+ programmes/dayGraduated 2026. Thesis: DetectIQ, a SAM3 + LoRA fine-tune for structural damage segmentation, taken to production.
3.2 / 4.0 GPAMy first taste of a real engineering org: courier-to-branch transfer automation and UX improvements on package-logistics microservices, built in pair programming under senior code review. Java Spring Boot, Docker, Jenkins, Datadog.
Led a two-person team taking an AI municipal call-center and CRM from zero to MVP in ~5 months: Django, PostgreSQL, GPT API, AWS Polly.
−80% TTS cost · 7s → 1.5s latency · 2 municipalitiesOptimized Node.js and PostgreSQL queries across 6 modules and rebuilt 5+ Flutter screens around reusable components.
~25% API response improvementCarbon-footprint tracking platform for 3 industrial clients with React and Node.js, plus SEO that grew peak traffic to 2,000+ weekly views.
1,000+ monthly emission recordsFour shipped products, each framed by the problem, the system built, and the measurable outcome.
A process-safety platform that turns HAZOP / LOPA methodology into structured, auditable risk software, with a hybrid-RAG layer grounded in plant documentation.
A personal operating system for tasks, time blocking, nutrition and workouts, with AI daily planning synced across web and mobile.
A decision-intelligence engine that tells you what to focus on now, and shows its reasoning. Not another task manager.
A market-entry strategy engine where 100+ specialized agents debate options, red-team assumptions, score surviving plans and assemble a board-ready playbook.
Agent tooling I use every day, sustainability, municipal AI and embodied AI.
A self-hostable job-search outreach system I built for my own search. It discovers companies, accepts an email only if it appears on the company's site and passes an MX check, drafts a personal message, then has a separate LLM call audit the draft for invented claims.
Nothing is sent without a human decision, and a bounce guard pauses new outreach once the bounce rate passes 6%.
Python stdlib · SQLite · BYOK LLM · GitHub Actions
One source of truth for 21 specialised AI subagents across 6 domains. Each role is written once and compiled into the native formats of Claude Code and OpenAI Codex, so the same reviewer, red team and implementer are available in both.
Node.js · Claude Code · OpenAI Codex
A carbon-footprint tracking platform and digital presence for an environmental consultancy, processing 1,000+ emission records every month.
React · Node.js
An AI municipal call-center and CRM, co-founded and built end to end: it routed citizen complaints, tracked cases, and reported KPIs for two real municipalities across 300+ complaint categories.
Reworking the request path and model calls cut spoken-response latency from ~7s to ~1.5s and TTS cost by ~80%.
Django · PostgreSQL · GPT API · AWS Polly
A small physical companion robot paired with a local-only reasoning core that builds memory and personality over time, designed to integrate with LifeOS and Octostra.
Status: product design and early prototyping. Next up — an on-device memory graph and a companion app that plugs into LifeOS and Octostra.
Local LLM · Memory engine · Robotics
My capstone fine-tunes Meta's SAM3 segmentation foundation model with LoRA to segment bridge and infrastructure damage — cracks, corrosion, spalling and 16 other classes — at the pixel level, on the DACL10K dataset. I trained just 0.25% of its 842M parameters, then shipped the model behind a FastAPI + React inspection platform: upload a photo, get back a colored damage mask with per-class coverage.
What I reach for, grouped by where it lives.
open to AI / software engineering roles where I can build and grow
Best fit: AI Product Engineer, Full-Stack AI Engineer, or product-minded software engineer roles on a team where I own a real problem end to end, ship it with the people around me, and keep growing alongside senior engineers.