nyami.fr

/ leadership

Leadership & delivery

What I shipped, and what it changed. Every figure below is documented in the linked case study; client-specific figures are never approximated.

Delivered agentic KYC corroboration workflows in a regulated environment

Source of Wealth / Source of Funds analysis combining structured banking data, external evidence and tool integration — LangChain/LangGraph orchestration with end-to-end traceability and human review, advanced to pre-production.

LangChain · LangGraph traceability ✓ human review ✓ pre-production ✓ Architecture Lab →

Delivered an autonomous multi-agent trading system under strict risk governance

Architecture separating probabilistic reasoning (18 AI agents) from deterministic execution — no order passes without the risk engine, hard kill switch has priority over everything. Paper validation before any real capital.

v1.5.0 shipped 608 tests 95% coverage kill switch ✓ AlphaPilot case study →

Led production LLM evaluation infrastructure in a regulated environment

From internal prototype to a platform the team can operate: enterprise SSO, reproducible GitOps deployments, auditable execution — delivered through the security gate with zero critical CVEs.

enterprise SSO ✓ prod-grade CI/CD ✓ 0 critical CVEs pre-production ✓ LLM evaluation case study →

Progressed enterprise conversational AI (Copilot Studio) from POC toward production

Secure authentication, controlled Power Automate ingestion, Dataverse integration within the existing governance frame, and reusable delivery patterns — in a regulated environment.

POC → production pathway enterprise authentication controlled ingestion reusable pattern Enterprise chatbot case study →

Directed a zero-downtime platform migration off a proprietary vendor

Multi-agency EU platform moved from Base44 to a Supabase + Cloudflare stack with progressive DNS cutover — no service interruption, no lost email, full stack ownership regained.

113 vendor refs → 0 zero downtime 7 tables + RLS 8 edge functions BPSI migration case study →

How I lead delivery

  • Evidence first: every claim on this site links to a case study, an ADR or a live dashboard — the same discipline I ask of teams.
  • Decisions written down: 16+ ADRs on this project alone; architecture reviews start from documents, not slides.
  • Quality gates in CI: eval scores (tool selection ≥ 0.80) and security scans (Trivy, Gitleaks, SBOM) block ship, not opinions.
  • Cost governance by design: budgets, quotas and kill switches are architecture, not afterthoughts — live proof on /eval.
  • Roadmaps published: every system states what ships next — and never claims what is not live yet.