About
I'm an AI Engineer and Senior Data Scientist with 4+ years of experience taking ML systems from problem scoping to production, across time series forecasting, computer vision and LLM automation. I hold a PhD in Physics from TU Dresden.
I enjoy working closely with customers — understanding their operations and economics, integrating with their data, and shipping practical, reliable AI that solves real business problems.
What I work on
Demand forecasting & decision optimisation
Sales and demand forecasting with DARTS, Nixtla and CHRONOS, plus newsvendor-style ordering that balances margin against stockout risk for each customer.
Computer vision
Food recognition with YOLO-family models and PyTorch, automated annotation with SAM and OpenCV, and real-time POS order matching.
LLMs & agentic workflows
Tool-calling LLM agents that automate internal analytics and reporting.
Production ML & MLOps
FastAPI microservices on Docker and Kubernetes, CI/CD, MLflow, and Sentry/PowerBI observability for near-zero downtime.
Apps
Interactive app · Python · Streamlit
Newsvendor Problem Visualiser
A graphical way to understand the newsvendor problem — the trade-off behind every daily order. Simulate demand, set stockout and overstock costs, choose a safety margin and watch the losses change; consecutive stockout days cost more, like real customer frustration. Then let the app search for the loss-minimising margin.
Experience
Nov 2025 – Present
Sr. Data Scientist
Demand Forecast Project Lead
- Lead the demand forecast project end to end, from customer scoping to features in production across 300+ locations.
- Built an in-house newsvendor-based ordering optimiser that tunes forecast quantiles to each customer's margin and stockout risk, minimising food waste without losing sales. Try the app ↗
- Improved sales forecasts by 20% during holiday and offer seasons.
- Automated POS integration for faster onboarding, and Sentry/PowerBI observability for negligible downtime.
- Built LLM tool-calling agents for internal analytics, and automated customer food waste reports (−50% manual effort).
Feb 2022 – Oct 2025
AI Engineer and Data Scientist
Time series forecasting — sales and demand
- Led AI-driven demand forecasting, reducing food waste by 15% and improving crew efficiency by 20%.
- Designed an automated forecasting pipeline using DARTS, cutting retraining time by 50%.
- Reduced query latency by 60% using FastAPI microservices with MongoDB, Docker and Kubernetes.
Computer vision — object detection and segmentation
- Deployed AI-based food recognition using YOLO-family models and PyTorch across 10+ restaurant chains.
- Built automated annotation pipelines with OpenCV and SAM, reducing labelling time by 50%.
- Delivered a POS order-matching system with 95%+ real-time validation accuracy.
Skills
AI & Deep Learning
GenAI / LLMs
Time Series Forecasting
Computer Vision
Software & Deployment
Data Engineering & MLOps
Analytics & Delivery
Languages
Education & publications
PhD, Physics — Technical University Dresden
Publication: Phys. Rev. B 103, 064425 (2021)
BSc–MSc, Physics — IISER Pune
Publication: Phys. Rev. B 95, 054401 (2017)
Data Science & Machine Learning Bootcamp — Le Wagon Berlin
Led project: restaurant location recommender (Python, TensorFlow, NLP, GCP, Streamlit)