FHAbout Farzad
AI & Time-Series Scientist | Operational Forecasting (Hydrology • Energy • Markets) | xAI • LSTM/Transformers • Azure MLOps
I turn complex time-series into reliable, explainable, deployable forecasts. In hydrology, I’ve delivered hourly 7-day (168-lead) streamflow forecasts for Swiss glacier basins with peak-event MAPE cut from ~11% to ~4%, and a regional 24-h hourly model for 40 flashy humid basins (Basque Country, Spain) with NSE ≈ 0.98. I also built a daily regional AI model for 531 U.S. catchments that outperforms all existing benchmarks.
What I do?
Operational forecasting systems: problem framing, KPIs, data audit, feature engineering, model design (LSTM/TCN/Transformers), HPO/ensembles, xAI/SHAP, uncertainty, backtesting, event-based verification.
MLOps on Azure/GPU: packaging, CI/CD, monitoring, alerts, retraining, documentation, handover.
Data engineering for forecasts: ingest/align multi-source forecasts (different horizons & resolutions), gap handling, ML-ready tensors.
Domains: rainfall–runoff/flood, humid and flash floods, glacier hydrology, early warning, plus energy load/renewables and market/demand forecasting.
Why clients book me?
Measured impact: proven event accuracy.
End-to-end delivery: from messy data to production pipelines your ops team can run.
Trust by design: xAI and uncertainty to support decisions and regulators.
Scientist + builder: 3× Journal of Hydrology (Q1), IEEE, Oral talks (AGU/EGU/HEPEX/IEEE/HydroML).
Stack:
Python (PyTorch, TensorFlow), Azure ML, Docker, Git, Conda; CI/CD & experiment tracking; metrics: MAPE/NSE/KGE/RMSE.
Availability
Open to EU/Swiss/UK/US/global remote; Autónomo in Spain; EU work-authorized. Will travel for short on-site engagements.
Engagements: fixed-scope projects or ongoing improvements/retainers/long-term.
Reason for low price: start of freelance work. In future I give discount to my older clients.
Lets start some today.
English
Native or bilingual
Spanish
Conversational
Persian
Native or bilingual
Can work on-site
Bilbao (up to 50km)
Experience
- Swiss water-engineering firm (2025)Independent AI Scientist (Freelance Consultant - Predictive AL Model Enhancement Project)TECHApril 2025 - October 2025 (6 months)Switzerland• Optimized and trained operational deep learning models for hourly streamflow forecasting (168hr‑lead).• Project scope: A known hydrologically complex glacier Swiss catchment• Deliverables (Written confirmed by client):1- Operational improvement of hindcast/forecast with MAPE >10% ≈ 4% at flood peaks2- Designed an operational pipeline on Azure ML with robust big data handling and model configuration.3- High-level recommendations on making research hindcast operative in real-time4- Specific configuration of AI models to operate in real-time based on field experiments5- Advised teams on DL design, hyperparameter tuning, and operational best practices for Glacier basins.
- Hydraulics Institute of Cantabria, University of CantabriaResearcher (Postdoc)TECHNovember 2020 - Today (5 years and 7 months)Spain- 2024-25: Researcher (Postdoc) — Hydraulics Institute of Cantabria, University of Cantabria • Spain• Deliverables under Operational Projects in Collaboration with Basque Water Agency (URA):1. Development of FEWS‑integrated adaptor connecting hydrological models to the old operational system.2. Multi-station/multi-catchment data engineering and dataset pipelines.3. Optimization and Re-Calibration of conventional hydrological models for Basque Country Catchments.• Deliverables under Research Line of Deep Learning in Hydrology:1. Advanced xAI in hydrology (SHAP, feature attribution) to improve model trust and adoption.2. High-accuracy optimized regional DLs across 531 USA catchments outperforming all benchmarks.3. Event‑based diagnostics and improvements by development of ensemble deep learning methods.4. Explainable AI to study the performance and learnings of regional AI models in the USA.5. Participation in development of research and operational proposals.6. High-level collaboration on Seagrass detection by AI models (Bio-team): method, codes, verification- 2020-24: Researcher (Predoc) — Hydraulics Institute of Cantabria, University of Cantabria • Spain• Deliverables under Operational Projects Collaboration with Basque Water Agency (URA):1. Calibration of conventional hydrological models for Basque Country Catchments.• Deliverables under Research Line of Deep Learning in Hydrology:1. Optimized DL models for regional rainfall‑runoff modeling with systematic hyperparameter optimization.2. High-accuracy AI/DL optimized regional models across 40 Basque catchments (NSE ≈ 0.98).3. Event‑based predictive improvements by development of ensemble deep learning methods.4. Designing advanced xAI method in hydrology (SHAP, feature attribution) to improve model trust.5. Input feature engineering and tensor design for AI model training/serving.
- —Central Office of Refah Chain Stores Co.•Data ScientistJanuary 2019 - October 2020 (1 year and 9 months)Tehran, Tehran Province, IranData Scientist in chain marketRefah is a multi branch chain supermarket like CarrefourI used to work as a data scientist there
Recommendations
Be the first to recommend Farzad
Help this freelancer shine by sharing your experience working together.
These freelancer profiles also match your criteria
Agatha Frydrych
Backend Java Software Engineer
4.7
(3)
2
Baptiste Duhen
Fullstack developer
4.6
(4)
5
Amed Hamou
Senior Lead Developer
4
(2)
7
Audrey Champion
Web developer
4.3
(3)
4
Education
- Ph.D. AI & Deep Learning for Timeseries ForecastingUniversity of Cantabria2025Ph.D.
- ML & AI for Environmental VariablesIHCantabria2023ML & AI for Environmental Variables