Job alerts
smir
full-stack engineering · applied AI

I build complete web products
— from interface
to production

Full-stack engineer with nine years of production experience. I deliver interfaces, APIs, data systems, automations and production infrastructure.
My work also covers reliable AI features and agents

software development9 years
team management4 years
scopeFull stack & AI
based inDijon
01

Complete web products

I handle the interface, API, database and deployment. I can take an idea all the way to production.

Claude CodeCursorTypeScriptReactNodePython
02

Useful AI features

I design agents, RAG search and structured AI workflows. Tests, evaluations and fallbacks keep them reliable.

ClaudeOpenAIRAGMCPFine-tuning
03

Reliable automation

I automate business work with n8n, APIs and queues. I include security, error handling and monitoring.

n8nVPSDockerOWASPWebhooks
/01selected systems

My latest projects

Four shipped projects. Each shows my role, the architecture and the result.

01 open-source infrastructure · active

LLM Failover Proxy

One endpoint that keeps working when an AI provider fails, rate-limits or slows down.

problem
Keep an OpenAI-compatible endpoint answering when a provider rate-limits, returns nothing usable or stalls mid-stream.
my ownership
Routing logic, failure classification, hedged requests, named model lists, CLI and terminal UI, persistence, tests and the npm release.
architecture
Node.js 22 behind one OpenAI-compatible HTTP and SSE contract, provider adapters, Promise.race hedging with AbortController, cooldowns persisted across restarts and zero runtime dependencies.
delivered
Pull request #4 merged upstream: 12 commits and +1,638 lines across 18 files, published as v1.8.0 on npm.

Forked from @ajoly21 and reviewed back into upstream, the shell in the agent lab below serves the same ordered model chain.

02 internal production platform

Extranet Fimainfo

One platform for telephony operations, reporting and an access-controlled AI assistant.

problem
Bring operational workflows and data from several services and SQL Server estates into one maintainable platform.
my ownership
Architecture and full-stack delivery across the web application, BFF, unified API, RAG service, BI and observability.
architecture
Next.js and Express over four SQL Server estates, FastAPI and Qdrant for permission-aware RAG, Docker, Traefik and OpenTelemetry.
delivered
Five services, around 60 operational screens and roughly 786 commits over 17 months.

Production and source code stay on the company network, the case study documents the architecture without exposing internal data.

03 live client platform

ASCoFaçade

A public website and four secure portals for a construction company.

problem
Replace disconnected marketing and site administration with a single role-aware platform.
my ownership
Product structure, UX, frontend, server actions, data model, access control, AI features and deployment.
architecture
Next.js 16, React 19 and Supabase with row-level security, resumable media uploads, document generation and provider-agnostic AI.
delivered
One public local-SEO site plus four secured role portals, in production.
04 live product

way2buy

A Telegram Mini App for sales, loyalty and cashback.

problem
Unify product discovery, loyalty, cashback, publishing and administration inside the channel customers already use.
my ownership
Product model, Telegram Mini App UX, backend, database, rules engine, scheduler, AI reports and serverless deployment.
architecture
Node.js, Express and SQLite with Telegram identity, bidirectional channel webhooks and idempotent scheduled jobs.
delivered
A zero-build client backed by 19 tables and 81 automated tests, available as a live Mini App.
/02experience

From IT operations to full-stack and AI

I moved from infrastructure to web products, then full stack and applied AI. I still use every layer.

2011—2017

Systems foundation

French Foreign Legion · IT operations

Hardware and software fleets, network security, hardening and operational L1/L2 support.

2017—2021

Web delivery

Képi Blanc · production web

Site components, editorial workflows, technical SEO, integrations and performance.

2022—2023

Product frontend

Novileads · remote

Responsive products from Figma, shared component work and browser/device quality assurance.

2023—today

Full-stack + applied AI

Fimainfo · production systems

Interfaces, APIs, data and delivery alongside agents, RAG, automation and production controls.

The result: I connect product choices to production limits, including cost, security and reliability.

/03working toolkit

What I can deliver and run

Four areas I can own from design to production.

engineering coverage

I work across the whole product.

The main skills first. The full technology list follows.

01

Product interfaces

React · TypeScript · accessibility · design systems

I turn business workflows into clear, responsive interfaces.

see evidence →
02

Services & integration

Node.js · Python · REST · webhooks · queues

I design APIs and integrations that handle errors, retries and third-party limits.

see evidence →
03

Data & delivery

PostgreSQL · SQL Server · Docker · CI/CD · observability

I manage databases, CI/CD, monitoring and production debugging.

see evidence →
04

Applied AI & automation

RAG · agents · evaluations · n8n · deterministic guardrails

I add AI features with clear tests, cost limits and fallbacks.

inspect the lab →
interactive technology wall
my stack is on the wall — move any note, double-click the paper to pin a tool I'm missing

swipe the wall · hold a note to move it
/04how the work gets done

Plan, build, review, run

The same delivery process for my code and AI-generated code. Nothing ships without review and a real test run.

Define the work first.

I write the goal, constraints, risks and acceptance criteria before implementation.

architecture · decomposition · NFR/SLO · risk list

Use agents. Own the decisions.

Agents can implement tasks. I decide what ships and what must be rewritten.

delegation · shared scratchpad · one owner per decision

Review with evidence.

Every acceptance criterion needs a test, code reference or command output.

score 0–100 · acceptance criteria · no "trust me"

Run the real stack.

I run migrations, backend, frontend and end-to-end tests before release.

db migrate · boot back + front · e2e · then ship

AI-assisted engineering system

My working AI agent lab

A live VPS lab where specialised agents plan, build, review and test work.

  1. 01input
    Youone message — Telegram or CLI
  2. 02orchestration
    Managerreads the task, picks the team, the model and the tooling
  3. 03routing
    Profileone team wakes up — Dev, SEO, Marketing, Security, Design
  4. 04execution
    Agentsplan → build → hand off through a shared scratchpad
  5. 05quality gate
    Gatea reviewer scores it and a verifier boots it — or it loops
  6. 06delivery
    Youworking software, with the evidence attached
interactive sandbox

Manager

The wire between me and every team. Give it a task — it tells you who it would wake up and why.

Run llmfp start in the shell above — your own proxy instance is what unlocks the agents.
/05personal system · job alerts

A live automation, end to end

A production workflow that collects, filters and reviews job listings on my VPS.

personal automation · production

My Job Alerts · workflow

A raw job page becomes a checked alert with a full history.

running on my VPS24/7
source ai agent filter database output
05minutes · polling cadence
03sources · monitored in parallel
4–8minutes · source to alert
0trigger
1sources
2scrape
3filter
4store
5act
swipe to explore the pipeline
live · a job moving through
at sources · LinkedIn
~6 min
avg. detection
3×
platforms unified
0
manual scrolling
24/7
running on my VPS
/06engineering foundations

The principles behind my decisions

Short notes on software, systems and AI principles I use in practice.

what I am looking for

A full-stack role.
Real products. Useful AI.

I fit teams that need one engineer across product, backend and production, with practical AI experience.

role focus
Full-stack engineer · AI engineer
based in
Dijon · France
work modes
Remote · EU time zones / hybrid · France
languages
French C2 · English B2 · Ukrainian / Russian native
start a conversation LinkedIn CV · EN CV · FR

Tell me what you need to build and what is blocking the team.