Climate & geospatial ML
Statistical downscaling, bias correction and forecasting on ERA5 / GCM-class datasets — U-Net architectures and custom likelihood losses, trained with reproducible xarray/NetCDF pipelines.
Available — remote & relocation
Hassan
Hassan
Kerroumi
AI EngineerFez, Morocco
I build machine-learning systems for climate and environmental data — and the backends, bots and automations that put them to work.
01 — About
Research-grade models,
Research-grade models,
production-grade systems.
AI engineering graduate of EIDIA, Euromed University of Fez. My research work sits in climate and geospatial deep learning; my product work is backend engineering and automation. I like problems where a model has to survive contact with real data and real users.
Backend & automation
Production REST APIs in Node.js/TypeScript and Python — webhooks, state machines, payments, notification templates, relational and NoSQL storage, Docker deployment, and n8n workflow automation.
Applied AI & LLMs
Computer vision, NLP, RAG systems and prompt design — plus the judgment to review AI-generated code for what it usually gets wrong: unhandled error paths, missing idempotency, unpersisted state.
PythonPyTorchTensorFlow
TypeScriptNode.jsFastify
Next.jsxarrayNetCDF
ERA5CMIP6GDAL
PostgreSQLMongoDBDocker
n8nOpenCV
RAGscikit-learnGit
02 — Experience
Where I've worked
Feb — Jul 20266 months
UM6P · IWRI
Climate Downscaling Research Intern
- Built an end-to-end deep-learning pipeline downscaling coarse GCM output (LMDZ, 35–250 km) to 0.1° precipitation and temperature fields over Morocco.
- Trained a U-Net model on ERA5 predictors and MSWEP observations, with a custom Bernoulli–Gamma loss for zero-inflated rainfall.
- Applied Scaled Distribution Mapping bias correction and ran inference on SSP245 / SSP585 emission scenarios.
- Engineered config-driven, reproducible geospatial pipelines with xarray and NetCDF.
Jul — Sep
Euromed Research Center
Research & Development Intern
- Research on cluster validity metrics under different normalization schemes.
- Built a mobile application assisting nursing students with diagnostic learning.
03 — Work
Selected projects
Research systems, shipped products, and the automation glue in between.
/ 01Research
Climate Downscaling for Morocco
Deep-learning statistical downscaling turning coarse global climate projections into high-resolution rainfall and temperature fields. Custom Bernoulli–Gamma loss for zero-inflated rainfall, feature-importance-driven predictor selection, distribution-mapping bias correction, and future-scenario inference.
/ 02● Live
SafarFlex
safarflex.comContributed to a Moroccan travel-booking marketplace connecting travellers with local partners — accommodation, transport, guided experiences and dining — with traveller–partner messaging, multilingual content and a three-sided traveller / partner / admin model.
/ 03Automation
WhatsApp Messaging Automation
An n8n workflow driving WhatsApp messaging through the WhatsApp API using WAHA as the gateway — session setup and authentication, inbound message events over webhooks, and outbound sends triggered by business events, with failures surfaced rather than swallowed.
Garbage Classification
CNN waste classifier at 92% accuracy, benchmarked against Random Forest and SVM.
Arabic OCR
CNN-BiLSTM architecture with an attention mechanism for Arabic text recognition.
News Classification
TF-IDF + SVM news categorisation pipeline reaching 98% accuracy.
Voice-to-Voice Translator
Real-time speech translation with GMM-based spoken-language identification.
ML Mobile Robot
End-to-end pipeline: sensor data collection, preprocessing, training and model comparison.
Image Filtering App
Image-processing application implementing a range of filter effects.
04 — Skills
The toolbox
Languages
PythonTypeScriptJavaScript
C / C++JavaSQL
Machine learning
PyTorchTensorFlowscikit-learn
U-NetCNN / BiLSTM
Climate & geospatial
xarrayNetCDFERA5
CMIP6GDALQGIS
Bias correction
Backend & data
Node.js / FastifyREST APIsWebhooks
PostgreSQLMongoDBNext.js
AI applications
LLM APIsRAGPrompt design
NLPOpenCVComputer vision
Infrastructure
DockerGitn8n
LinuxLaTeX
05 — Background
Education & languages
Education
AI Engineering Degree
EIDIA — Euromed University of Fez · 2021–2026
International Baccalaureate
Physical Science · 2020–2021
Languages
ArabicNative
EnglishB2+
FrenchIntermediate
SpanishBasics
Certificates
Machine & Deep Learning
Natural Language Processing
Generative Adversarial Networks
Mathematics & Statistics
SQL — MySQL
06 — Contact
Let's build something
Open to roles and projects in applied AI, climate & environmental data, and backend engineering. Email is the fastest way to reach me.