About

Hi, I’m Mert, a full-stack ML engineer.

I love trying to understand the structure, dynamics, interactions, and evolution of things. I’ve spent the last decade building ML systems end to end, across startups and scale-ups.

Some of what I’ve built:

At Tenderd (Principal ML Engineer, 2024–now), an AI platform for equipment and vehicle operations: vision-based crane cycle detection and productivity estimation, driver monitoring, worksite hazard detection, and telemetry-based equipment utilization, fuel, and emissions modeling.

At BECO Capital (Lead Data Scientist, 2022–2024), a leading MENA VC, where I was the first data scientist: greenfield data and cloud infrastructure, an in-house deal-flow platform, probabilistic entity resolution across fragmented data sources, and an internal knowledge engine integrating dealflow, transcripts, and market data.

At Swvl (Senior ML Engineer, 2020–2022), a NASDAQ-listed mass-transit unicorn: real-time demand-responsive pricing for marketplace balancing, origin-destination demand modeling, transit network design, and demand-side modeling for churn, route propensity, and demand forecasting.

At Careem (Data Scientist, 2019–2020), the region’s top ride-hailing platform (acquired by Uber): metric-agnostic time-series forecasting, customer-lifetime-value and price-elasticity modeling, and a growth-experimentation platform.

At Trendyol (Data Scientist, 2017–2019), Turkey’s largest e-commerce company (acquired by Alibaba), where I was part of the founding data science team: dynamic pricing, demand forecasting, product ranking, and predictive segmentation.

The trouble with summing a person up like this is that it files them under headings, which flattens more than it captures. So take the following as gestures, not categories, though these, at least, are non-negotiable and will outlast most things about me. I’m a cat person. I’m drawn to history, literature, and anthropology. I hike and snorkel. I might add more, but only if the data backs it up for long enough. This paragraph took me an unreasonable thirty minutes, so the probability of change, over any reasonable horizon, is near zero. But never zero.