Ask IA
Wassim SalmiAI Engineer

I build products that put AI to work, in production.

AI Engineer, full-time. I build AI into products, like on this portfolio: RAG, agents, document extraction, plugged into the client's data and tools. Python and FastAPI on the AI side, TypeScript, NestJS and Next.js for the rest. I've been coding since 2020, and every day with Claude Code, which I follow closely so I can code well and fast. Available now, open to international roles.

Day to day, in production
↓ Ask me a question

Ask me anything

Ask a question or pick a suggestion

How I got here

From Dofus to my first contract in production, in eight steps.

From 2008 to today, eight turning points. Hover over a year.

I've been on a computer since I was six. I'd wander around the internet looking for some info or a game, and I figured out how the web worked very quickly. I started coding in 2020, and little by little I swapped almost all my gaming time for coding time. To me it's the same thing, except it's more useful.

Today I build AI into products: RAG, agents, document extraction, plugged into the client's data and tools. Python and FastAPI when the AI calls for it, and around that, TypeScript: NestJS and PostgreSQL on the back end, React and Next.js on the front end, React Native when it's mobile.

What matters to me is taking a step back. When you're asked for a feature, there's the code, but there's also the time you have and what the company can afford. A lot of developers over-optimize. Optimizing costs time, and time is a resource the company doesn't always have. So first I look at what needs to change, why, how far it goes, what it implies. Then I execute.

Writing the code you're asked for is half the job. The other half is seeing what it will cost in six months.

The technical foundation, grouped by area. Every piece is here because it was used on a shipped project, not to make the list longer.

Applied AI

Existing models, plugged into a real product

Front end

What the user has in front of them

Back end & data

The API, the model, what handles the load

Cloud & production

Getting it online, and keeping it online

Only five per block: my real stack goes well beyond this.

How I work

Coding fast with agents, yes. But coding well.

Vibe coding feels fast, but you build and break things at the same time. With my agents it's just as fast, but nothing gets merged without unit tests, e2e tests and a PR review. I call this loop 3M.

  1. The product, broken down
    Macro
    The product, broken down

    Sprints, user stories and a dependency graph.

  2. The waves
    Meso
    The waves

    Agents in parallel, each in its own worktree.

  3. The user story
    Micro
    The user story

    Four agents, one role each.

The four agents

  1. Planner
    Planner

    Breaks the story into tasks

  2. Implementer
    Implementer

    Writes the code, not the tests

  3. Tester
    Tester

    Writes the tests, never touches the code

  4. Reviewer
    Reviewer

    Merges, or sends it back

With Claude Code, one Git worktree per agent, Jest and Playwright.

The cloud, beneath the code

My master's specialization. A developer who understands their deployment, their network and their database makes better decisions, and knows how to ship something to production that holds up.

Infrastructure as code

Infra that's written, versioned, replayable. No server set up by hand.

Containers

The same image locally and in production.

AWS

On the job at Linkpick: Cognito for authentication, S3 for files.

Going to production

A CI that tests before deploying, a Linux VPS behind nginx.

Projects
01 / 05

Where I've worked, and on what

Five roles, from bootcamp to missions in production.

  1. AI Engineer ·

    2025 - present
    My personal projects banner · AI products
    • AI Tutor: a tutor that answers learners from their school's courses and nothing else. RAG with OpenAI embeddings and pgvector in PostgreSQL, explicit refusal off-curriculum.
    • MyCommentAI: a Chrome extension that comments on LinkedIn in the user's own tone. FastAPI API, RAG over a vector database, GPT to write. Published on the Chrome Web Store.
    • AI Prospecting: seven specialized agents built with Hermes on a Linux VPS, from monitoring to writing follow-ups. 268 qualified targets in 16 days, sending stays human.
    PythonFastAPINext.jsRAGOpenAIpgvectorPlaywrightDocker
  2. Fullstack Mobile Developer ·

    Oct 2025 - Sep 2026
    Freelance mission · 12 months
    • Production management SaaS for dough-based trades, on web and mobile: production sheets, recipes, sourdough starter, stock, equipment.
    • Design of the business calculation engine, based on trade rules validated with the baker before being written.
    • NestJS / Prisma / PostgreSQL API, Next.js front end and Expo app: the calculations are shared between web and mobile, and covered by tests.
    • Internationalization in twelve languages with no hard-coded strings, and subscriptions handled by Stripe on the web and RevenueCat on mobile.
    React NativeExpoNext.jsNestJSPrismaPostgreSQL
  3. Fullstack Developer & Architecture ·

    Jan - May 2026
    Freelance mission · 4 months
    • End-to-end architecture of a secure LMS: NestJS, Next.js, PostgreSQL, Docker.
    • Fine-grained access control (RBAC) checked on the API side, per resource.
    • CI/CD pipelines with automated tests (Jest) run before every release to production.
    NestJSNext.jsPostgreSQLDockerJest
  4. Fullstack Developer ·

    2025
    Freelance mission
    • Online ordering site for an Asian restaurant: menu, cart, Stripe payment, delivery or pickup.
    • Full back office and kitchen screen: the restaurant changes its menu, opening hours, delivery zones and promotions without going through a developer.
    • Next.js and React front end, Zustand cart and forms validated with Zod; Express API with Sequelize on MySQL, JWT authentication.
    • Deployed in Docker containers behind nginx, product images on Cloudinary and confirmation emails through Resend.
    Next.jsReactExpressSequelizeMySQLStripe
  5. Software Engineer Fullstack ·

    Oct 2024 - Mar 2026
    Work-study · 1 year 6 months
    • Development across the platform's 16 Java Spring Boot microservices, and three features designed and delivered on my own end to end: class schedule, absences, QR code attendance.
    • Design of the AI module: CV extraction (PDF and OCR) into schema-constrained JSON with Mistral, and classification of job offers into ROME, France's official job taxonomy. In Python, with FastAPI.
    • Automation module built on my own: asynchronous architecture with Celery and Redis, one independent worker per task and real-time status tracking.
    • React then Angular front end across the four user spaces, with a code organization by feature domain that I proposed. Docker containerization, AWS Cognito and S3.
    • Proposed migrating the back end to NestJS, approved by the team and started before I left.
    Java Spring BootAngularReactPythonFastAPIDockerAWS
  6. Technical Instructor · La Capsule

    Mar - Aug 2024
    Intensive bootcamp · 6 months
    • Trained more than 20 developers on the TypeScript stack: React, Node.js, SQL.
    • Coaching on Clean Code, Git Flow and good review practices.
    ReactNode.jsSQLGit Flow
Education
  1. Information Systems Architect, Cloud track · Epitech

    2026
    Master's level
  2. Web & Mobile Application Designer & Developer · La Capsule

    2024
    Bachelor's level
  3. Web & Mobile Web Developer · AFPA

    2022
    2-year degree

What I'm looking for, concretely

The role, the setting, availability. No fluff.

Role
Role
AI Engineer, full-time
Setting
Setting
Startups, scale-ups and software companies
Scope
Scope
From design to production
Location
Location
Open to international roles · on-site, hybrid or remote
Availability
Availability
Immediate
Languages
Languages
Native French, professional English, Arabic (currently learning)

What they say

Four written client reviews, translated from French.

YacineFounder · MyKnead
Working with Wassim means having a real project partner, not just a contractor. He took the time to understand my business, he guided me in my choices and he always told me things straight, including when it wasn't what I wanted to hear. The result: a project that moves forward on solid foundations.
ArnaudCTO · Linkpick
Organized, reliable and productive, you can trust him with any mission and he will pull it off brilliantly.
LiaDirector · Club des Office Managers
He was very responsive, efficient and fully respected the specifications I had given him. He even spotted several inconsistencies in them and pointed them out to me right away.
MohamedCo-manager · ServiceAsian
Beyond his solid technical skills, he stood out for his teaching, taking the time to explain the different steps to us and helping us get to grips with the tools.

Got a role to fill?

Tell me the context: the product, the current stack, the team size, how much on-site time, the deadlines. I reply within 24 hours. The CV is just below, and references are available on request.

Or write to me

If booking a slot feels too early, the form works too.

A question before committing thirty minutes, or no slot that works for you? Write here, it goes straight to my inbox. Otherwise, the address is still wassimassalmi@gmail.com.

Wassim Salmi, AI Engineer