AI Engineer · AI products, data platforms and cloud infrastructure
I build production AI systems and data platforms that keep running.
Remote · Canadian and Jordanian citizen
- Product delivery: At eeva, I co-led a platform for document-grounded answers, persistent context and controlled service workflows. Seven capabilities I personally built were adopted by the production team and shipped.
- Memory and integrations: I was the primary engineer on the standalone memory service, handling relationships, changing facts, versioning and retrieval. I owned voice and messaging integrations and deployed two MCP servers on Cloud Run.
- Evaluation: I built the evaluation harness that compared four retrieval configurations in an April 2026 internal experiment. Category reports exposed weaknesses behind aggregate scores and informed representation and ranking choices.
Explore the AI platform and memory service
- Acquisition: I built 160 source collectors and acquired 370,000 documents from manufacturer sites and public archives, using a shared runtime with provenance, deduplication, retries and cost controls.
- Document processing: I designed and operated the eight-stage processing platform, including extraction choices, source-grounded identity checks, targeted recovery and publication review. The production-clean catalog held 110,000+ manuals in September 2026; structured outputs preserved document identity and source references for retrieval.
- Cloud foundation: As eeva's only cloud engineer, I owned environments, infrastructure as code, access, deployments and migrations across Azure and Google Cloud. I created staging and modernised development and production.
- Operating responsibility: I built shared scheduling, execution records and monitoring for collection workloads. I led production database migration recovery and resolved a separate failover fault with Microsoft support, turning both incidents into monitoring rules and runbooks.
Cloud foundation · Operating platform
Production delivery and independent builds across AI, data and cloud, with implementation evidence linked throughout:
- Agents & integrations: Multi-step orchestration, typed tools, MCP, structured outputs and approval boundaries, with WhatsApp and LINE messaging integrations. Google ADK, LangGraph, FastAPI, OpenAI, Claude / Anthropic and Gemini.
- Memory & retrieval: Standalone memory services, changing facts, retained history and retrieval. PostgreSQL, pgvector, RAG and configuration experiments.
- Evaluation & model selection: Authored harnesses, adversarial tests, Recall@k/MRR, model routing and spending controls. Agent Cost Down uses NVIDIA Nemotron through Nebius; its reports retain scoring and latency qualifications.
- Data engineering: Acquisition, extraction, targeted recovery, provenance, reconciliation and quality-controlled release. Python, SQL, pandas, DuckDB, Azure Functions, Blob Storage, Pub/Sub, Dataflow and BigQuery.
- Applied ML & scientific data: Scientific-data preparation, reproducible baselines, leakage-aware splits and bounded analysis tools in independent NASA-data projects. Python, scikit-learn, pvlib and OpenAI.
- Software & cloud delivery: Python APIs, TypeScript interfaces, tests and database-backed services; Azure, Google Cloud, Docker, Terraform, Bicep and GitHub Actions. Delivery, access, migration and recovery.
Explore the capabilities and interactive explanations
BEng in Electrical and Software Engineering, McGill University.
Claude Certified Architect - Foundations, Anthropic. The certification covers agentic architecture, tool and MCP design, Claude Code workflows, structured output and context management.