Skip to content
lapoule.dev
fr Contact
News All news
Consulting & development · AI-enabled

Handcrafted code,
at the speed of AI.

Coding data architect for long-term engagements, and fixed-price projects from €5,000 excl. VAT, all included — delivered fast, delivered right, documented to software-engineering standards. XP, test-first, ML / RL at the core.

Start a project Our method
Throughput
×5
Fixed-price project
from €5,000 excl. VAT
Standards
XP / TDD
Our services

Two ways to work together, one standard.

Whether you need lasting reinforcement or a turnkey project, you get the same level of craftsmanship — accelerated by AI.

Time & materials · long term

Coding data architect, embedded in your team

A data architect who codes: designing your platforms — modelling, pipelines, streaming, ML — and building them hands-on, augmented by AI. An architect's vision, a senior developer's throughput.

  • Data architecture: models, pipelines, streaming, governance
  • Hands-on: the architecture ships as code, not slides
  • Human-AI pair programming, test-first, continuous review
  • Part of your team: Slack, stand-ups, product ownership
Day rate depending on scope and duration Discuss an engagement
Fixed price · from €5,000 excl. VAT

Your project delivered, in as little as 4 weeks

A tightly scoped project, delivered quickly, done right, documented to software-engineering standards. One price, one date, one deliverable. No surprise change orders.

  • Scope defined together in 30 minutes
  • Continuous delivery on Cloudflare previews
  • Code, tests, docs, runbook and monitoring included
  • Full ownership transfer on delivery
From €5,000 excl. VAT — quote after scoping Scope my project
Method · in motion

The XP cycle, accelerated by AI

Red → Green → Refactor → Ship. On every turn of the loop, the AI proposes and the senior engineer decides. Tests are the contract: written before the code, still green after.

Method

XP, test-first, cost-aware.

At lapoule.dev we build to Extreme Programming standards: short iterations, user stories, test first — without ever losing sight of the economics. A test, a feature, a pound of code: each one is justified by the value it creates.

User stories

No endless specification document. Every need is phrased as "As a … I want … so that …", with a testable acceptance criterion and an explicit economic criterion.

Short iterations

One-week sprints, weekly demos, continuous releases. You see the product move forward every Friday, not every quarter.

Test first

We write the expectation before the implementation. AI speeds up the build but never waives the contract: measured coverage, mutation testing where the value justifies it.

Human-AI pairing

The consultant doesn't outsource their brain to AI: they use it as a pair-programming partner. Generate, critique, refactor. The senior engineer stays accountable for the result.

Economic criterion

Every task is weighed by expected value versus cost. We don't build what doesn't pay off. We don't test what doesn't break. We refactor what slows us down.

Full-stack data

Your data doesn't sleep in silos. ML, AI and reinforcement learning are at the core of the product, not a cosmetic layer bolted on at the end.
Stack · in motion

Data flows through the stack, all the way to the edge

Python or Rust for processing, TypeScript for interfaces, Cloudflare on the way out. Latency below the psychological threshold, predictable infrastructure costs, multi-region by default.

Stack

The technologies our consultants master — and recommend.

Our consultants work across two circles of languages, two circles of cloud, and a front-end stack chosen deliberately for productivity and SEO. No hype, no dogma: choices that hold up technically and economically.

Languages · first circle

  • Python data, AI, critical scripts
  • Rust latency, reliability, edge
  • TypeScript Astro front end + React islands

Second circle

  • Java enterprise integration
  • Scala data pipelines, streaming

Cloud · first circle

  • Cloudflare Workers, R2, D1, Queues, Vectorize

Second circle

  • AWS EKS, Lambda, S3, SageMaker
  • GCP BigQuery, Vertex AI, GKE

Front end · our own stack

  • Astro SSR server rendering, multi-page, SEO
  • React (islands) targeted interactivity, no wasted JS
  • TypeScript strict mode everywhere

What you're reading runs on this stack. The source code is the demo.

Tech radar · beta

What we adopt, trial, assess or hold: 36 technologies and practices, drawn from production engagements.

See the radar
In-house projects
  • Docx System docx.lapoule.dev

    An MCP server that plugs Claude, Cursor or any AI agent straight into Microsoft Word.

  • Dead or Alive dead-or-alive.ai

    Pay-per-hire external referrals: anyone can hunt the right profiles for the right companies, on top of living CVs.
Learning · in motion

The loop that learns, in production

Spec → Build → Ship → Measure → Learn → Decide. Every deployment is a signal, every signal feeds the next decision. Reinforcement isn't a module: it's the system.

Ready to hatch your next project?

One hour to scope. As little as four weeks to deliver. A lifetime of tests to never regress.

Fixed price — from €5,000 excl. VAT Time & materials