Data scientist · Berlin
10 years turning messy data into decisions that hold up under scrutiny. I work on incrementality, uplift, and lifecycle analytics professionally — this is where I explore the rest.
Projects
An end-to-end uplift modelling pipeline on synthetic food-delivery data, using a T-Learner with CatBoost to find which customers a voucher actually saves — not just who's likely to churn. Constructs ground-truth causal effects to validate the approach, then adds budget optimisation and profit-curve analysis showing how churn-based targeting underperforms uplift-based selection.
A growing collection of AI skills for feature engineering — self-contained tools an LLM can invoke to reason about and construct features for common data science problems. Built in public, platform-agnostic across Claude, OpenAI, and Gemini. Current skill: customer inactivity prediction. Next: uplift modelling.
Writing
Knowing who might respond to a voucher is only half the problem — the other half is knowing where to stop. On using unit economics to find the profit-maximising cutoff, beyond which targeting more customers just burns budget.
Why voucher targeting needs uplift modelling, and what the four quadrants actually mean. On the difference between attrition risk and causal responsiveness — and why conflating the two wastes budget on customers who would have acted regardless.
Published on the Just Eat Takeaway tech blog. How JET's in-house Customer Data Platform consolidates ~3 TB of data into unified 360° customer profiles — powering segmentation, journey tracking, and self-service data access for marketing and product teams, with GDPR compliance built in across every market.
Published in Analytics Vidhya. A walkthrough of building time series forecasting models for stock prices, covering modelling and analysis.
A deep dive into Spark's accumulator API — when the built-in types aren't enough and how to extend them for custom aggregation needs.
About
I'm a data scientist based in Berlin with a background in causal inference, uplift modeling, and lifecycle analytics. I work at Just Eat Takeaway, where I own the incrementality and attribution frameworks that shape how we allocate marketing spend across tens of millions of customers.
Outside work I'm building in public — mainly tools at the intersection of LLMs and data science. Currently: a multi-provider feature engineering skill for churn and inactivity prediction, exploring what happens when you let different models reason about the same feature space.
I also hold a Post Graduate Diploma in Applied Statistics from the Indian Statistical Institute and spent my first three years in tech as a software engineer.
Hire me
Open to senior analytics and data science roles in Berlin — especially where causal inference is a real differentiator, not a buzzword. I care about shipping work that changes decisions, not dashboards that get ignored.