Alexander El-HajjData Scientist · GenAI · Montreal QC · (01)
Data scientist & generative-AI builder, based in Montreal.
I like hard problems and the unglamorous engineering that makes AI actually work. Day job: modernizing secure data systems at Statistics Canada. The rest of the time: building, breaking, and shipping GenAI. Eight years turning research and messy data into things that run.
I research and chase down hard problems, then build the thing that unblocks the team. I'm happiest understanding a system end-to-end and finding the simple version of it.
For eight years I've worked at the seam where data meets reality — secure public-sector analytics at Statistics Canada, applied ML and OCR research for federal partners, and reinforcement-learning work that ended up in a Springer volume. More recently that curiosity pulled me into generative AI: retrieval over messy document stores, fine-tuning small models under tight constraints, on-device inference with real safety boundaries, and agents on cloud infrastructure. I care about evaluation, observability, and the unglamorous integration work — the parts that decide whether a prototype survives. I hold a Reliability Status security clearance.
Now · 2021—present
Senior Analyst - Survey of Household Spending (SHS)
Statistics Canada · Hybrid
Co-leading the migration of production survey workflows from legacy SAS to open-source R and Python inside a secure government environment. I build validation pipelines, internal analytics tools, peer-review workflows, and the GitLab practices that help teams ship reliable systems with less manual drag.
2022—2023
Generative AI Engineer
Independent Consultant · Remote
Built custom AI-video pipelines for client deliverables, including work for Wunderman Thompson and Standard Chartered Bank. Configured GPU environments, extended open-source diffusion/video tooling, and turned fast-moving research into repeatable production workflows — some of it passing 500K views.
2020—2021
Data Scientist — Data Science Division
Statistics Canada · Ottawa
Co-developed Project Cyclops, a Python OCR pipeline that helped Health Canada field agents check product-label compliance from phone photos. Co-led agent-based reinforcement-learning research for PHAC on COVID-19 mitigation — later published in Springer.
2018—2021
Analyst & Research Lead - Longitudinal and International Study of Adults (LISA)
Statistics Canada · Ottawa
Led survey-release workflows and cross-divisional coordination, ran technical workshops, and co-authored two research papers on food insecurity — including validating the Food Security Module. Founded StatCan's first Kaggle competitive group.
(03)
Selected work
Shipped & recognized
01
Special Recognition · Mila AI Safety Hackathon
GuardAI a stacked guardrail that protects youth in digital crisis spaces
Built with team MindCraft. Keyword filters miss youth crisis because risk is a trajectory — it hides in coded language and multi-turn subtext. We fine-tuned a bilingual (EN/FR) EuroBERT-610M classifier on 85,000+ synthetic samples from a six-phase data-generation pipeline, using two-stage curriculum learning across 23 risk categories. A fast ~24ms classifier flags obvious signals; gray-zone cases escalate to a cascaded LLM judge. We tuned for recall on purpose — a false alarm gets filtered by a counsellor, a missed crisis does not.
Kept private under hackathon and safety constraints. Happy to walk through the pipeline, training tricks, and evaluation in detail.
EuroBERT · PyTorch · Curriculum learning · Synthetic data · 8-bit Adam · BF16 · Cascaded LLM judge
03
Shipped · 2026
Gemma Flares on-device health agent with a CI eval gate & hard safety boundary
A local-first Flutter/iOS app (Gemma via LiteRT-LM) for tracking IBD flare patterns. Deterministic code owns all risk math, routing, and persistence; the on-device model is sandboxed to explaining grounded evidence — it never computes risk or suggests medication. The interesting part is the operational layer: a persona-driven eval gate runs in CI on every PR — it scores each agent turn against safety and RAG-grounding contracts, flags rag_used_when_forbidden violations, and blocks the release if any hard-safety check fails. A runtime benchmark service captures cold-start and P50/P95 on-device generation latency.
persona-eval suite hard-safety pass P50/P95 latency115 tests · adversarial prompt-injection suite
Flutter · Dart · Gemma · LiteRT-LM · SQLCipher · CI eval gate · Persona suite · Latency benchmarks · Adversarial tests
02
Shipped · 2026
Local RAG Engine source-grounded retrieval over 20GB+ of PDFs
A retrieval system for document libraries too large to hold in memory. Resumable, idempotent S3 ingestion with multiprocessing and size guards; Qdrant + FastEmbed with MMR retrieval; answers grounded in their sources with an explicit "I don't know" refusal when the context doesn't cover the question. Ships with health-check and vector-store audit scripts so you can see what the index is actually doing.
Momentum an SMS-first personal finance agent on GCP
Syncs bank transactions through Plaid, classifies spending with Gemini on Vertex AI, and runs budget workflows over Cloud Functions, Cloud Run, and Pub/Sub. Handles the unglamorous part properly: idempotent pending-to-posted transaction state.
An in-progress design for agent infrastructure — tool use, task routing, and repeatable local/cloud workflows, with the operational layer (MCP-style integration, tracing, prompt contracts) as the focus. Early-stage; the production patterns here are proven in Gemma Flares's shipped agent + eval gate.
Multi-agent systems · MCP · ReAct · Tracing
06
Ongoing
Agents Learning Track HF Agents · DeepLearning.AI GenAI with LLMs
Sharpening the implementation patterns behind production GenAI: agent design, evaluation, LLM application architecture, and responsible deployment.
Hugging Face · DeepLearning.AI · Evaluation
Academic ML
07
Coursework
Heart Disease Prediction supervised vs. unsupervised, interpretable
Compared algorithms for predicting heart-disease outcomes, with the emphasis on model selection, feature behavior, and interpretable evaluation rather than a single accuracy number.
Steganalysis detecting hidden data in digital media
Researched steganalysis methods for spotting concealed information in images — the statistical signals, the image features that give it away, and the digital-forensics use cases.
The same stubbornness that debugs a pipeline at 1am shows up here too. I train hard, I'm in the mountains when I can be, and I cook like it's a build process.
Muay Thai & BoxingTraining with the legend SaenchaiSkiingWhitewater, Nelson, BCHikingPedra da Gávea, BrazilAlwaysBuilding side projects & devouring papers
