
Kamil D
AI Engineer
Skills

See my services

Work experience
AI Agent Systems Engineer | Research Automation
Self Employed • Freelance
Jul 2026 - Present • 2 mos
Designed and built Pressroom, a production-oriented multi-agent platform that automates research, verification, writing and publishing while keeping human control over critical decisions. • Architected a stateful LangGraph workflow with specialist agents for research, source analysis, thesis development, adversarial review, drafting, factual auditing, quality assessment and publishing. • Created a hybrid control system combining LLM reasoning with deterministic Python checks for citation traceability, unsupported numerical claims, inconsistent figures, broken sources, repetition and content duplication. • Implemented PostgreSQL checkpointing, retry policies, bounded repair loops, cost ceilings and restart recovery for reliable long-running agent workflows. • Added human-in-the-loop approval using workflow interrupts, with configurable review and publishing controls. • Built source-quality classification, research evidence ledgers, claim-level citations and validation against original webpages. • Implemented model routing and token and cost tracking across different agents and tasks. • Added duplicate-topic detection using embeddings and vector similarity. • Built an operational console showing the live agent graph, execution status, artifacts, events, costs, review queues and publishing decisions. • Connected approved content to a version-controlled static publishing system used by CostOfEntry.com. • Containerized the application with Docker and PostgreSQL/pgvector and configured Linux and nginx deployment. The project demonstrates my ability to design dependable AI agent systems that go beyond simple prompt chains by adding workflow state, evidence controls, evaluation, recovery and operational visibility.
Confidential
Full-time • 5 yrs
AI and Digital Transformation Consultant
Jan 2022 - Present • 4 yrs 8 mos
AI and digital transformation consultant with more than four years of experience supporting banks and insurance companies in Morocco. • Led discovery workshops with business and technology stakeholders to understand operational challenges, customer journeys and transformation priorities. • Mapped business processes, identified automation opportunities and translated business needs into clear functional and technical requirements. • Contributed to solution design, project scoping, delivery planning and coordination between business teams, developers and decision-makers. • Worked in regulated financial environments where traceability, security, data governance and controlled implementation were essential. • Applied structured consulting techniques including stakeholder interviews, process mapping, requirements analysis, gap analysis, prioritization, acceptance criteria and delivery documentation. • Helped teams evaluate how data, automation and AI could improve internal operations, customer service and decision-making. This experience allows me to connect AI engineering with real business requirements instead of treating AI as an isolated technical feature.
AI Engineer and Full-Stack AI Product Developer
Apr 2026 - Aug 2026 • 4 mos
I design and build production-ready AI applications from initial architecture through deployment. My work focuses on multilingual RAG, AI agents, knowledge retrieval, research automation and analytics products. Selected systems: • 9anoun: Built a French and Arabic legal knowledge assistant using Moroccan legal codes and court decisions. Implemented PDF ingestion, multilingual embeddings, dense vector search, BM25, hybrid retrieval, reciprocal-rank fusion, cross-encoder reranking, query rewriting and cited answers. • Pressroom: Architected a stateful LangGraph research and publishing system with specialist agents, source verification, factual audits, deterministic quality checks, repair loops, cost controls, checkpoints, human approval and recovery after interruptions. • Social Agent: Built a SaaS platform that transforms YouTube comments into multilingual audience intelligence, evidence-linked recommendations, dashboards and PDF reports. Implemented asynchronous jobs, Redis queues, retries, accounts, workspaces and delivery services. • Visual Test: Created browser automation for detecting contrast, layout, touch-target, focus, console, request and visual regression problems across multiple viewports. Technical methods include RAG, prompt architecture, structured outputs, grounding, hallucination controls, retrieval evaluation, model routing, asynchronous processing, streaming APIs, automated testing and containerized deployment.