Tacklab is my practice. I assemble the right mix of human and agentic developers to take a product from first sketch to live, monitored production.
AI-native CTPO (Chief Technical & Product Officer). I also take on CTO-as-a-Service engagements and startup consulting.
The right team of people and agents — design, build, deploy, and keep it healthy in production.
$ engagements scale to the need — from a 3–5 day prototype to test an idea, to a full production build I maintain and evolve after deployment
I build real applications end to end — from concept and architecture through to a deployed, monitored product. AI-native throughout, blending human and agentic developers to move fast without sacrificing quality.
Embedded or on-demand technical leadership: architecture, roadmap, build-vs-buy, team topology and delivery. The judgment of a CTO, sized to where your company actually is.
Technical strategy for founders: AI adoption, engineering transformation, and the hard technical decisions — grounded in shipping, not slideware.
I don't ship and disappear. I run what I build — monitoring, reliability, quality tracking and steady iteration on live systems once real users depend on them.
Real systems I designed, built, and took all the way to production.
An AI voice agent that answers the phone for veterinary clinics — in French.
Nolia answers the phone for veterinary clinics in France and Quebec: it books, reschedules and cancels appointments, looks up patients, and triages emergencies — in natural spoken French, accents included.
I built the whole thing. The real-time voice pipeline (Twilio into an ElevenLabs STT → LLM → TTS cascade tuned for sub-600ms turns). A code-orchestrated decision core that keeps business rules in code rather than trusting the prompt — so the agent can only book a slot it actually offered. Multi-tenant integrations into the software clinics already use (GmVet, Google Calendar, Vetup). It runs on AWS, and I operate it in production — tracking conversation quality, tool-call reliability, latency and recognition accuracy on the hard parts: proper names and phone numbers.
Measuring how brands show up when people ask an LLM instead of a search engine.
GEO — Generative Engine Optimization — measures how a brand appears when people ask ChatGPT, Claude, Gemini and Perplexity: the visibility traditional SEO simply can't see. Tacklab GEO is the product this site originally launched to sell. Today it stands as proof of what I can deliver end to end.
I designed and built the entire platform: a multi-pass LLM analysis pipeline that separates an unbiased response from entity extraction and sentiment scoring, a multi-tenant data architecture on GCP and BigQuery, automated daily analysis with intelligent scheduling, and the dashboards on top. Production-grade end to end — CI/CD with path-filtered deploys, feature flags, smoke tests and instant rollback across roughly 140K lines of TypeScript, Vue and Python.
A 90-day post-merger integration plan and a 6–12 month AI-native engineering transformation for a €30M+ spend-management SaaS merging two products and two engineering cultures. Platform and data architecture, org design (Team Topologies, inverse Conway), and an AI governance model that turns compliance into a moat. A compact showcase of how I think as a CTO.
Ongoing 1-on-1 coaching for the founding CTO of an early-stage startup in factory process automation — a sparring partner on architecture, technical decisions and engineering leadership as the company grows. CTO-as-a-Service in its most direct form: helping a technical founder make better calls, faster.
What lets a small team deliver end to end is the way I work: human and agentic developers run together, with custom tooling I've built to keep the agents fast and safe.
A code-review system where independent agents check for bugs, guideline compliance and git-history regressions, score every finding by confidence, and filter the noise — sized to the change, so a five-line fix doesn't get an architecture lecture.
Reusable, specialized tools for the repetitive parts — secure database access through SSM tunnels, browser automation for testing, docs and diagram generation, clean session handoffs. The boring work, automated.
Agents are team members with explicit budgets, guardrails and feedback loops — not a black box. Business logic lives in code, agents accelerate the work, and a human stays accountable for every merge.
// packaging some of these tools to open-source soon
> I take on a small number of projects at a time. Tell me what you're working on.