Hi, I'm Kubilay Yavuz
|
I build production AI systems, agentic workflows, and data products. My recent public work is a UK property intelligence stack: typed public-data clients, geospatial tools, an AVM layer, and MCP servers for Rightmove, Zoopla, and OnTheMarket.
Recent GitHub work: agentic UK property intelligence
Python monorepo · Updated Apr 24, 2026
uk-property-intel
Eight-package UK property intelligence workspace covering portal parsers, public API clients, geospatial tools, AVM modelling, and a natural-language LangGraph agent.
Public MCP servers · Created Apr 18, 2026
Rightmove / Zoopla / OnTheMarket MCP
Three installable MCP servers exposing normalized property search, listing detail extraction, branch profiles, safe inquiry actions, and SQLite listing-change firehoses.
Government + open-data clients
Public-data intelligence layer
Typed async clients for EPC, HMLR Price Paid, planning, flood, Companies House, ONS, council tax, air quality, geology, radon, procurement, and postcode data.
Who I Am
Senior Data Scientist building production AI
I'm a Senior Data Scientist at McKinsey & Company based in London. I specialise in deploying LLM-powered applications, agentic workflows, and machine learning systems end-to-end — with a strong focus on real-world delivery and client impact.
Beyond consulting, I co-founded QApture Ops — a production AI platform for automated mobile game QA using multimodal agentic AI (OpenCV, YOLO, PaddleOCR, GPT, Claude). I also co-founded Eminence Games, publishing mobile games and apps on iOS/Android.
I hold a BSc in Industrial Engineering from Bogazici University, ranked in the top 0.01% nationally. I combine mathematical rigour with engineering pragmatism to ship things that work.
Bogazici University
BSc Industrial Engineering · 2015–2020
Top 0.01% · National University Entrance Exam
Production AI Systems
LLM workflows, agentic pipelines, computer vision — deployed end-to-end, not just notebooks.
Real Business Impact
118% margin improvement, 70% hour reduction, 20% pick-pack efficiency gains — at McKinsey, Getir, Infarm.
Full-Stack AI Delivery
From FastAPI backends and Celery workers to Next.js dashboards — I own the full stack.
10+ Hackathon Awards
ING, SAP, Turkish Airlines, Euroleague — proven record of building fast under pressure.
Professional Experience
- Built an LLM workflow to interpret legacy property deeds and store structured outputs in PostgreSQL. Led 2 engineers to deliver backend and appraiser dashboard.
- Led end-to-end data platform migration for 2 clients — medallion architecture, Power BI dashboards for business users. Led team of 4 engineers through development and deployment.
- Built elasticity-based pricing engine transitioned to production — contributing to 118% quarterly realised margin improvement.
- Led geospatial analytics, footprint optimisation, M&A opportunity modelling, and mixed-integer optimisation for cost reduction across retail and PE clients.
Entrepreneurial Experience
QApture Ops
Co-founder · 2025 – Present
Production AI platform for automated QA testing of mobile games. Multimodal agentic workflow using OpenCV/YOLO, PaddleOCR, GPT/Gemini on Android/iOS. FastAPI + Celery + Supabase + Docker on GCP. Built the customer-facing Next.js dashboard.
Eminence Games
Co-founder · 2024 – Present
Designed and published multiple mobile games and apps on iOS/Android. Responsible for gameplay systems, analytics, and economy design.
TripSorter
Builder · 2024
AI-powered travel planning app generating dynamic, multi-day itineraries using LLM agentic workflows with external tools for routing, weather, and local logistics.
Technical Expertise
AI / ML
LLMs, agentic workflows, computer vision, classical ML
Backend & Infrastructure
APIs, distributed systems, cloud, MLOps
Data Engineering
SQL, pipelines, warehousing, BI tools
Frontend & Other
Web, mobile, game development
From mathematical modelling to production deployment — I work across the full AI stack, with a focus on shipping systems that generate measurable business impact.
Featured Projects
Current build
UK Property Intel
Built a typed UK property intelligence workspace spanning portal parsers, government API clients, geospatial primitives, valuation models, and a LangGraph natural-language agent.
Current build
UK Property MCP Servers
Built three MCP servers exposing normalized listing search, detail extraction, estate-agent branch profiles, safe contact actions, and listing-change firehoses.
Selected case study
QApture Ops — AI-Powered Game QA Platform
Co-founded and built a production AI platform that autonomously plays and tests mobile games using computer vision, OCR, and LLM-powered decision-making.
Selected case study
Teknofest AI Competition
Participated in Teknofest AI Competition, achieving 3rd place in semi-finals and 5th in finals with a vehicle and pedestrian detection model.
Selected case study
Production Pricing Systems
Developed and productionised multiple pricing engines combining machine learning, operations research, and Bayesian approaches.
Selected case study
Warehouse Optimisation System
Built a warehouse optimisation system using mixed-integer programming and TSP routing applied across 7 warehouses.
Selected case study
Turkish Airlines Travel Datathon
Developed predictive models for passenger counts on two-legged flights using a dataset of 93 million rows.
Selected case study
LLM-Based Deed Extraction System
Built an LLM workflow to interpret legacy property deeds and store structured outputs, leading 2 engineers to deliver backend and appraiser dashboard.
Get In Touch
Location
London, United Kingdom