Dean Shabi
AI engineer and two-time founder. I've spent seven years putting machine learning into production, mostly forecasting for energy markets, and built the AI behind Otty and Katalo. I now own the solar forecasting stack at Renewcast.
Experience
2025 to now
Italy · Remote
Senior Data Scientist · Renewcast
- Own the solar forecasting stack. Cut portfolio forecast error from 15% to 6% in five months by pairing plant physics with a learned correction.
- Built one model per fleet that corrects the next four hours from live plant data. It beat persistence on 98% of solar plant-months in backtests.
- Replaced hand-picked model releases with a statistical test against the forecasts customers received, documented in four ADRs.
Machine learning · Weather prediction · Python · API development
May 2025 to now
Remote
Founder · Otty
- Built a career agent that finds roles, judges fit and applies for candidates over WhatsApp, within rules they approve.
- Kept judgment and permission apart. The model decides what fits, and code re-checks the candidate's rules before anything is submitted.
- Cut cost per agent turn 5x by fixing prompt-cache routing, loading tools only when needed and capping stale context.
- Moved the product across three agent runtimes in four months without changing what the agent is allowed to do.
AI agents · TypeScript · Postgres · Product
Feb 2026 to Sep 2026
Remote
Co-Founder · Katalo
- Built an image-editing pipeline where a calibrated vision judge decides what ships and its rejections become repair prompts.
- Built a shared queue and adaptive rate limiter for four AI providers. Capacity leases expire, so a crashed worker frees its slot.
- Released prompts and judge versions the way code ships, with snapshots, hashes and a staging step.
LLM evals · Convex · TypeScript · Image models
2024 to May 2025
London · Remote
Founding Data Scientist · tem.
- Led the AI behind RED, tem.'s main product for renewable energy.
- Built Rosso, a pricing engine that trades portfolio risk off against growth.
- Built half-hourly forecasts with horizons of several years.
AWS · Python · PyTorch · dbt
2023 to 2024
Prague & London · Remote
Data Scientist for Energy · AmpX
- Built time series models for generation, load and market price forecasting.
- Built battery degradation models for hundreds of assets.
- Set up the company's MLOps platform on AWS and Kubernetes, from training to monitoring.
AWS · Kubernetes · Airflow · Grafana
2019 to 2022
Prague
Data Scientist · Datamole AI
- Delivered ML projects end to end for manufacturing, automotive and agritech clients.
- Worked on predictive maintenance, anomaly detection and time series forecasting.
- Built data pipelines for robot, sensor and IIoT data.
SQL · Docker · Python · Azure
2014 to 2019
Israel
Project Lead · Captain · Israeli Air Force
- Led engineering teams on large technology projects for F-16 and F-15 fighter jets.
- Ran RF and signal processing R&D with Israeli defence companies.
- Built ML models for computer vision and data analysis.
Python · TensorFlow · MATLAB
Education
2018 to 2019
MSc Artificial Intelligence
Technion, Israel Institute of Technology · 2018 to 2019
2010 to 2014
BSc Electrical and Electronics Engineering
Tel Aviv University · 2010 to 2014
Electro-optics, control engineering and bioengineering.
2010 to 2014
BSc Physics
Tel Aviv University · 2010 to 2014
Astrophysics and relativity.
Skills
- Data science and ML
- Machine learning, Time series forecasting, Deep learning, NLP, Computer vision, Optimisation
- Programming
- Python, SQL, R, MATLAB, JavaScript, React, FastAPI
- Cloud and DevOps
- AWS, Azure, Kubernetes, Docker, MLflow, Airflow, dbt
- Libraries
- TensorFlow, PyTorch, scikit-learn, Pandas, Keras, Streamlit