Build a first AI product
You have an idea or a concrete use case. I turn it into a working product — from design through production deployment, with a clear path from prototype to live use.
Product-minded AI engineer · Builder · Operator
GenAI from buzzword
to production.
I've been shipping production GenAI since the early days of the field. I understand your problem, advise on the approach, and build the solution: writing the code, deploying to production, and training your team to run the system with confidence. You work directly with me.
GenAI Pioneer · in production since 2021
8 years Google & Microsoft
GenAI pioneer since 2021
3 continents
Full-stack · ultra-autonomous
scroll
01 / What I do
WHAT I ACTUALLY DO
Audit & unblock an AI initiative
Your team started, got stuck, or is going in circles. I diagnose what's blocking you and build a realistic action plan to get things moving again.
Train teams & executives
Your engineers want to build reliable GenAI systems. Your leadership wants to understand what they're deciding on. I train both through practical workshops teams can apply the next day.
Ship to production faster
You're at 80% but the last mile keeps slipping. I step in to close the gap and get what needs to ship actually shipped, on time.
Services
One offer, three ways to engage.
Most requested
I build your AI system
Full-stack GenAI system, production-grade. From problem definition to deployed product.
- Full-stack GenAI, production-grade
- From problem definition to deployed product
- Proprietary code and fully transparent architecture
- Typical: 4–12 weeks, solo or small team
I audit your AI strategy
20+ stakeholder interviews, gap analysis, prioritized roadmap.
- 20+ stakeholder interviews
- Gap analysis and prioritization
- Deliverable: written report + exec presentation
- Typical: 2–4 weeks
I upskill your teams
Hands-on workshops, prompt engineering, LLM evals, adoption strategy.
- Hands-on workshops and prompt engineering
- LLM evals and adoption strategy
- From 1 team to 250+ people
- Typical: 1 day to 3 months
02 / Work
WHERE IT SHIPS
Recently trusted by
Ellipse Animation
AI tools embedded in pre-prod → post-prod
Full confidentiality on proprietary assets
Problem
Animation studios need AI that respects confidentiality, fits the artist's creative flow, and plugs directly into the production pipeline.
Approach
Built AI tools complementing production workflows end-to-end: AI-coding stack, asset cataloguing, image/video studio capabilities tuned for confidentiality, domain-specific AI agents scoped to Ellipse's workflows.
Impact
Searchable asset catalogues, studio workflows aligned with artistic direction, agents operating within Ellipse's own domains.
Pre → postWorkflow span
AgentsDomain-specific
AI codingStack foundation
Special Jury · Sagas · CB News 2026
Brut
AI Studio for editorial visual production
150 daily users · Jury award
Problem
Producing original historical documentaries at media cadence is incompatible with archive-only footage.
Approach
Designed an internal AI Studio letting journalists generate, edit and adapt visuals themselves — under full editorial control. Rolled out across newsrooms on 3 continents.
Impact
Documentaries shipped weekly with a fraction of the production team. Used daily by 150 people across Brut Africa, Brut India, Brut US.
150Daily users
3Continents
Special JurySagas · CB News 2026
Reference available
Silver 360° · CB News 2026AI for Efficiency · DGE · 2025
Brut
Internal AI suite — Brut Radar & Brut Scan
85% adoption · 4.7/5 NPS · Prix DGE 2025
Problem
Tens of thousands of audience comments and editorial signals need structuring at media cadence.
Approach
Architected and shipped Brut Radar and Brut Scan — LLM-augmented journalistic monitoring and audience comment triage at scale. Audience insight also productised for brand partners (e.g. Groupe ADP).
Impact
Brut Radar and Brut Scan became load-bearing newsroom tools — not experiments. An 85% adoption rate and a 4.7/5 NPS reflect real, sustained daily utility. The video-to-SEO pipeline earned Brut the AI for Efficiency prize from France's Direction Générale des Entreprises.
85%Adoption
4.7/5NPS
Prix DGE2025
Reference available
Les Echos
Maia — deploying internal AI to a flagship newsroom
Institution-wide deployment · Change management
Problem
A leading business daily rolled out Maia (internal AI assistant) and needed real editorial adoption with measurable day-to-day usage.
Approach
Hands-on deployment partner: change management with editorial teams, use-case mapping per desk, prompt engineering, integration with journalistic workflows, training cohorts.
Impact
Maia becomes part of the desk — what started as an IT rollout became a genuine journalistic capability shift.
Multi-deskRollout scope
EditorialGrade adoption
TrustedAI ops partner
ESCP × OpenAI
ChatGPT Edu rollout — institution-wide AI literacy
80+ champions trained · 250+ people upskilled
Problem
A top European business school wanted to embed AI capability across every function through a rollout that turns training into day-to-day practice.
Approach
In lockstep with OpenAI — owned adoption of ChatGPT Edu across all staff; trained 80+ cross-functional champions; built a champion network per department.
Impact
A self-sustaining internal AI community shipping concrete projects and building capability across the institution.
ChatGPT EduSchool-wide deploy
80+Champions trained
All staffFaculty · IT · Ops · Mktg
Reference available
Pathé
Pathé AI programme — upskilling, adoption & automation
23 teams · 250+ people · Automated slides
Problem
Pathé needed shared AI fluency across teams and practical automation for recurring reporting workflows.
Approach
Bespoke training for 23 teams (250+ people): hands-on sessions tailored to each role, from distribution to communications. Built automated presentation workflows (PowerPoint and Google Slides) — turning a recurring manual task into a repeatable process. Advisory on advanced AI use cases tied to audience forecasting and release strategy.
Impact
Lasting lift in day-to-day AI leverage at Pathé; reusable deck automation; clear path from training to concrete use cases.
23Teams trained
250+People upskilled
PPT & SlidesAutomated
Mozza × Seyna
Seyna GenAI roadmap — audit, prioritisation & 4-agent test bed
20+ interviews · 4-agent POC · 3-week prototype → prod
Problem
A leading French insurtech broker needed to map GenAI opportunities while keeping the existing product roadmap moving.
Approach
Freelance AI engineer for Mozza, on-site with Seyna. AI audit from 20+ interviews; narrowed 7 ideas to 2 high-ROI bets; co-built a sandbox with 4 agents (RAG for unpaid rent & health, commercial health comparison, document extraction).
Impact
Published Mozza × Seyna case study; Seyna continued prioritised tracks with clear technical guidance.
20+Interviews conducted
7 → 2Prioritised uses
4Agents built
Reference available
Awards
AI for Efficiency
Direction Générale des Entreprises
AI Action Summit · Paris · Feb 2025
Special Jury · Sagas
Grand Prix CB News Créativité IA & Data
Brut · AI Studio
Silver 360° · Radar & Scan
Grand Prix CB News Créativité IA & Data
Brut · Radar & Scan
03 / Use Cases
USE CASES
MEDIA & NEWS
Your journalists produce 10× more content than your team can process.
GenAI pipeline that ingests, structures, and formats video + text content at editorial speed. Live in weeks.
ANIMATION STUDIO
You need AI tools that fit your pipeline and protect your IP.
Custom GenAI integration built around your creative workflow. IP-safe and pipeline-native.
SCALE-UP · AUDIT
You've tested 5 AI tools and now need one path that reaches production.
AI maturity audit and prioritised roadmap. You leave with a clear, prioritised build plan your team can execute.
C-SUITE · TRAINING
Your leadership team is making high-stakes GenAI decisions and needs a clear operating model.
Half-day or full-day session. Decision-makers leave with a working mental model and a vendor filter they can use the next morning.
TECH TEAM · TRAINING
Your engineers know the APIs and want a system design practice for reliable GenAI.
Hands-on workshop: evals, prompt engineering, RAG architecture, agent patterns. Real code, real problems.
ENTERPRISE · ARCHITECTURE
You need a GenAI architecture that your team can own after I leave.
System design, stack selection, documentation, handoff. Your team takes over with clear operating context and real ownership.
INSTITUTIONS
You need independent technical expertise to evaluate AI systems or suppliers.
Evaluation frameworks, vendor assessment, compliance-aware AI audit. Experience: European Commission level.
STARTUPS · BUILD
You need a GenAI feature in production before your next funding round.
Full-stack build: design, code, deploy, iterate. You work directly with me, with fast decisions and continuous execution.
04 / About
ALEXANDRE LAVALLÉE
One person. From strategy to production. I understand, I advise, I execute. I write the code, I deploy, I debug — and I train your team to keep going.
8 years at Google and Microsoft across EMEA and APAC. Today: AI Engineer at Brut and Ellipse Animation Studio. Detached AI Expert at Bpifrance Le Hub. AI Evaluator for the European Commission. Serial founder in AI: Selas Studio, Passeport IA.
I've been building production GenAI since 2021, from the field's early production wave. Full-stack: I design, architect, code, deploy, and train. Media & Entertainment anchor: editorial tempo, creative pipelines, IP constraints.
I work directly with clients. I write the code myself and own the result end to end.
Technical stack
Languages
PythonTypeScriptSQLBash
AI Engineering
Claude CodeCursorMCP ServersSub-agents & SkillsLangGraphLLM EvalsSelf-hosted LLMs
Stack & Cloud
ReactNext.jsFastAPIGCPAzureAWSDockerOn-prem VMs
RAG in production: 5 mistakes I've seen everywhereChoosing your LLM in 2026Claude Code in production: my AI engineer workflow
ALEXANDRE LAVALLÉE
Product-minded AI engineer
05 / Process
PROCESS
Discovery
30 min
We talk. I understand your context, constraints, and what actually needs to ship.
Audit
1 week
Audit, architecture review, or scoping. You get a clear picture of where you are and concrete next steps.
Build
sprint-based
I design, code, deploy. Weekly updates. You know exactly where your project stands at every step.
Handoff
always included
Documentation, training, handoff. Your team leaves with clear operating context, the right habits, and real ownership.
06 / Common questions
FREQUENTLY ASKED QUESTIONS
What do you do exactly?
I'm a product-minded AI engineer — I build, audit, and train. Concretely: I design and ship production GenAI systems, run AI audits that turn into prioritised roadmaps, and deliver hands-on workshops for technical teams and executives. From strategy to deployed product, you work directly with me.
How do we start working together?
A 30-minute discovery call is enough to understand your context and confirm we're a fit. From there I scope the work, we align on next steps, and we build. You reach me directly from the first message, and I reply personally.
Do you build in production or just advise?
I write the code, deploy to production, debug what breaks, and train your team to run the system with confidence. This is hands-on engineering — not slide decks. My engagement ends when your system is live and your team owns it.
What kind of projects fit best?
GenAI systems that need to reach production: RAG pipelines, autonomous agents, LLM evaluations, internal AI rollouts, and first AI products built from scratch. I also run diagnostics for teams that started and got stuck. If it needs to ship, it's in scope.
Do you work remotely? What's your technical stack?
Remote-first — the vast majority of my engagements run entirely remotely, with the same level of quality and speed. Core stack: Python, TypeScript, Next.js, FastAPI, LangGraph. Cloud: GCP, Azure, and AWS. On-site is available for workshops or kick-off phases in Paris.
How can teams get trained on GenAI?
I run bespoke hands-on workshops for both technical teams and C-suite. Engineers get practical sessions on evals, prompt engineering, RAG architecture, and agent patterns. Leaders get a working mental model and a decision framework they can apply the next morning. From a single session to a multi-month upskilling programme.
Articles
Latest articles
27 August 2026
What a GenAI project costs in 2026: budget, timeline, ROI
A practical 2026 guide to GenAI project cost, timeline, ROI, scoping, POCs, production delivery, run costs, and Build vs Audit vs Train decisions.
Read article22 August 2026
RAG in production: the 7 mistakes that break quality (and how to fix them)
The 7 RAG mistakes I fix in production: chunking, retrieval evals, reranking, metadata filters, hallucination guardrails, latency, cost, and monitoring.
Read article18 August 2026
Choosing your LLM in 2026: GPT, Claude, Gemini, open-source — how I decide
My practical framework for choosing GPT, Claude, Gemini or open-source models in production: task fit, latency, cost, evals, context and EU constraints.
Read article07 / Let's work
Bring me the problem that matters — I'll come back with a working system and a clear plan.
An email or LinkedIn message is enough. Tell me what you're building, what feels blocked, or where you need a senior hand, and I'll reply personally.