What I've led, delivered, built.
The three projects on the home page are only an extract. This is the whole body of work, grouped by the role I actually held: the projects I led, the ones I delivered as paid client work, and the ones I built on my own.
Back to the home pageProjects I led
Engagements run at ETIC INSA Technologies where my job was to steer the project rather than build it: prospecting, scoping, pricing, coordinating the development team and managing the client relationship.
%20--%3e%3csvg%20version='1.1'%20xmlns='http://www.w3.org/2000/svg'%20xmlns:xlink='http://www.w3.org/1999/xlink'%20x='0px'%20y='0px'%20viewBox='0%200%20200.05%20200.05'%20style='enable-background:new%200%200%20200.05%20200.05;'%20xml:space='preserve'%3e%3cstyle%20type='text/css'%3e%20.st0{fill:%23E60019;}%20.st1{fill:%23FFFFFF;}%20%3c/style%3e%3cg%20id='Calque_1'%3e%3c/g%3e%3cg%20id='_x32_.2'%3e%3crect%20class='st0'%20width='200.05'%20height='200.05'/%3e%3cg%3e%3cpath%20class='st1'%20d='M59.33,113.55c11.94,0,22.08-6.55,23.93-7.75c1.28,2.24,6.36,7.75,17.35,7.75c8.7,0,18.35-4.33,21.57-6.41%20c1.97,2.64,6.01,6.42,18.33,6.42c8.14,0,13.3-2.82,16.54-6.71c0.06,2.05,0.14,4.04,0.23,5.57l4.93,0.76%20c-0.16-5.14-0.03-22.94-0.03-23.21c0-22.36-17-23.94-20.56-23.94c-5.53,0-10.86,1.15-16.47,3.36l-0.4,5.89%20c2.18-1.02,9.18-4.03,16.16-4.03c5.49,0,9.72,1.84,12.59,5.48c-18.9,9.94-30.18,17.94-31.88,24.57%20c-4.84,3.27-13.36,7.05-20.95,7.05c-8.08,0-11.56-3.49-13.07-5.52c20.74-13.95,31.49-22,30.17-28.36%20c-1.18-5.73-9.7-8.42-17.34-8.42c-7.81,0-13.18,3.37-15.72,5.99c-4.08,4.21-6.03,10.45-5.78,18.55c0.11,3.54,0.81,7.23,1.98,10.46%20c-3.18,2.07-11.85,7.31-21.8,7.31c-11.82,0-16.68-9.23-16.68-18.55c0-9.03,4.54-18.56,16.68-18.56c6.83,0,13.68,2.88,16.02,3.96%20l-0.39-5.88c-4.41-1.72-10.21-3.28-15.23-3.28c-20.1,0-21.76,16.87-21.66,23.74C37.76,96.68,39.6,113.55,59.33,113.55z%20M155.87,81.15c0.64,1.61,1.18,3.69,1.51,5.65c0.02,0.1,1.38,10.56-4.19,16.95c-2.88,3.31-7.22,4.76-12.73,4.8%20c-12.57,0.11-13.79-4.66-13.79-4.66C125.36,100.97,129.6,95.07,155.87,81.15z%20M87.92,75.86c2.88-3.31,7.15-4.98,12.66-4.98%20c5.94,0,10.93,2.07,11.44,4.45c0.38,1.78-1.3,4.06-8.98,10.22c-0.08,0.07-7.84,6.65-17.86,12.6c-0.63-1.6-1.12-3.4-1.45-5.34%20C83.71,92.7,82.35,82.25,87.92,75.86z'/%3e%3crect%20x='42.73'%20y='129.15'%20class='st1'%20width='114.58'%20height='4.88'/%3e%3c/g%3e%3c/g%3e%3c/svg%3e)
Intelligent requirement extractionProject leadETIC INSA Technologies
Client: CEA, the French Alternative Energies and Atomic Energy Commission. Opening a strategic account around automated document analysis.
Prospecting, qualifying the need, technical scoping, pricing, coordinating development, running the project and managing the client.
A local, sovereign AI pipeline to industrialise requirement extraction from large, inconsistent specification documents, with structured and traceable exports.
- OCR
- NLP
- LLM
- HolIAGen / Open WebUI
- LangFlow
- Python
Project leadETIC INSA Technologies
Client: PROVEXI, a firm specialised in technical inspection and industrial document management. Automating the handling of very different documents: inspection reports, invoices, quotes, field reports.
Qualifying the need, functional scoping and steering the development of a modular Intelligent Document Processing solution.
A chain combining automatic document classification, OCR where needed and configurable no-code extraction: the business teams adjust the extraction rules, the classification and the prompts themselves, with no developer involved.
- IDP
- OCR
- NLP
- Streamlit
- JSON / Excel export
Project leadETIC INSA Technologies
Client: ATEXO. Exploring an AI module to automate the analysis of public tender files and shortlist applications.
Qualifying the need, technical scoping, pricing, steering the AI prototype, coordinating development and managing the client.
Proof of feasibility, and an architecture ready to industrialise.
- RAG
- LangChain
- FastAPI
- PostgreSQL
- Docker
- OVH Cloud
Client work I delivered
Work built with my own hands for a real client, under a paid engagement: concrete deliverables, clear objectives, a client with something at stake.
Technical consultantETIC INSA Technologies
Client: ARICI, a structural works SME in Bayonne. Every public tender means reading through a pack of several hundred pages to find what actually binds the firm, keeping only the structural works package, spotting the contradictions, then writing a technical proposal that matches the scoring criteria. Delivered and signed off in September 2026.
Design, architecture and full-stack development of the application, deployed inside the client's own infrastructure (native Windows, no Docker).
The application ingests the tender pack, classifies it, extracts and deduplicates the requirements, filters by package, flags contradictions, pulls out the scoring criteria, produces the analysis report and the response structure, then drafts the technical proposal from a vectorised library of the firm's own material. Word and ZIP deliverables.
- FastAPI
- Python
- Celery
- PostgreSQL + pgvector
- React 19
- Mistral
ConsultantETIC INSA Technologies
Client: SOCOTEC. A data and BI engagement comparing internal commercial performance against market dynamics, for the group's executive committee.
Designing the tool, harmonising the data, data modelling and building the Power BI dashboard from the Sitadel 2 API and the internal data lake.
A multi-level dashboard by département, business segment and branch.
- Power BI
- Sitadel 2 API
- Data lake
- ETL
- Star schema
Things I built alone
Projects carried on my own, with no client brief behind them: ideas, experiments, curiosity.
Design and developmentInternal tool
INSA Lyon's student-run consultancy. Writing commercial proposals by hand took too long and produced inconsistencies between the CRM, the contract and the invoicing.
An internal tool that automates 95 % of proposal production across the different service lines: scope, methodology, budget and payment schedule. After the client's feedback the proposal can still be edited by hand, then turned into a contract automatically, with the budget, schedule and invoicing data written into HubSpot.
The whole chain brought down from 4–6 hours to 10–30 minutes (reported figure).
- Python
- REST API
- HubSpot API
- LLM
- Docker
- Nginx
A local automation, designed end to endSelf-taught
A personal project built during an internship at Parkhouse Bell, a talent research firm based in London. Producing client deliverables meant re-typing LinkedIn profiles into Excel and then into PowerPoint, by hand.
After working out what the platform allowed, an approach built on the CV PDF that LinkedIn itself offers, so the extraction stays within permitted use, then a local Python script producing consolidated Excel exports and automated PowerPoint decks.
- Python
- pandas
- python-pptx
- openpyxl
- lxml
IoT wildfire detection systemSensor network design and alert visualisationAcademic project
An academic project at the end of the foundation years at INSA Lyon: design a system to prevent and detect wildfires.
A network of connected sensors measuring temperature, air and soil humidity, wind, CO₂, particulates and GPS position, with a view onto the alerts.
Adaptive transmission and low power draw: the data rate follows the risk level.
- IoT
- LoRaWAN
- ESP32
- HTML / CSS
- JavaScript
