📍 Barcelona, Spain
I've spent 8+ years building backend systems, and the last few taking AI out of the notebook and into production. Right now I'm at Revolut, where I delivered an internal AI assistant end to end — with MCP tool integrations — serving both internal staff and Revolut People users.
- 🤖 I work on agentic workflows, RAG, vector search and MCP tooling
- ⚙️ Backed by years of Python, FastAPI, Kafka and microservices at scale
- 🧪 Before that: productionising ML models and maintaining MLOps pipelines
- 🌍 Arabic (native) · English (professional) · Spanish (basic)
- 🔗 abdullahalsaidi16.github.io
Trains elderly people to recognise phone fraud through consented AI mock scam calls,
then gives their family a debrief. Consent is modelled as a single state machine, and only
the category of disclosed data is ever stored — never the value.
FastAPI Twilio SIP React 19 Cloud Run
→ Live site · GitLab
🔥 FUSION
Ranks 14,002 hexagons across Spain by what would be lost if they burned — fusing ESA
WorldCover, Kontur population, OpenStreetMap and 523 weeks of ERA5 weather into one weekly
priority score. Smarter caching cut API calls from 280 down to 30.
H3 rasterio Shapely
→ Live site · Repo
🔒 Maskode
A privacy-first fork of the OpenCode agent. A local model strips names, emails and
phone numbers out of your files before anything reaches an LLM provider — then restores
them in the response.
local inference PII redaction
→ Repo
A talk demo showing that an agent is just a ~60-line loop, and that one @tool decorator
can serve both an agent and an MCP server without duplicating schemas.
| Role | When | |
|---|---|---|
| Revolut | Software Engineer | Jul 2024 → now |
| Glovo | Software Engineer | Nov 2022 → Jul 2024 |
| Oneragtime | Backend Developer | Oct 2021 → Nov 2022 |
| STOWK | Backend Developer | Oct 2019 → Oct 2021 |
| L-One Systems | ML Engineer | Nov 2018 → Dec 2019 |
🎓 MSc Data Science, Harbour.Space · BEng Computer Engineering, Damascus University
