Rebuilding a legacy business platform without stopping the engine.

Context

SpotInfo’s platform is the company: it is where construction project data is processed, qualified and delivered to paying clients. That platform, built in ASP in 2009 and never maintained since, was reaching end-of-life: crashes were becoming more frequent under growing volume, response times were slow, and its dated ergonomics scared new clients away during demos.

The client asked me to lead the full rebuild with three goals: usability, stability and speed — and one absolute requirement: lose none of 15 years of historical data.

My role

Full project leadership — team recruitment, architecture, planning, supervision of development in close collaboration with the client, data-migration strategy, then driving the introduction of AI into the production flows.

What was delivered

The new platform

• Interface redesigned from scratch, oriented toward simplicity: usability difficulty was the number-one reason new clients dropped off;
• Next.js frontend, Express (Node.js/TypeScript) backend replacing the monolithic ASP;
• Performance transformed: platform 10× faster than the previous one;
• Stability restored: crashes gone, continuous updates without interruption.

The data migration

• 50,000 projects and all their dependencies (stakeholders, destinations, works) — around 200,000 records;
• extraction from the legacy SQL Server, transformation to the new cleaned model, loading and consistency checks;
• 15 years of history preserved — the company’s number-one asset.

AI integration (phase 2)

Once the platform was stable, we mapped every manual task and integrated AI flow by flow: automatic project qualification from architect sites (monitored in real time), automatic analysis of provided PDFs and websites, automatic processing of town-hall emails. Python orchestration with the Gemini and ChatGPT APIs.

Structural decisions

Recruit the team, don’t just manage it

The project started with hiring: four profiles sourced from my Cameroonian network, selected for their complementarity (design, backend, frontend, fullstack). Building your own team commits you: its wins and losses are yours.

Client collaboration as a living specification

The domain — construction-project life cycles, stakeholders, key contact moments — was documented nowhere: the knowledge lived in the founder’s head and in the ASP code. Instead of an endless specification phase, I set up close, continuous collaboration: frequent demos, iterative validation, short adjustments. The client co-built their platform.

A/B testing with real users before the cut-over

For two months the real users worked on the new platform in parallel with the old one. Every friction, every business-logic gap, every bug reported was fixed before the final switch — turning the riskiest moment of the project into a non-event.

AI introduced task by task, with humans reskilled

Production automation was not a big bang: each flow was automated, verified, then rolled out. The processing team went from 5 to 2 people, refocused on verification and quality control — since data is the company’s asset, humans remain the final guarantors of its reliability.

The challenges met

Rebuilding without stopping the engine. The legacy platform, however fragile, generated daily revenue. Parallel construction, repeatable and tested migration, a 2-month double-run: business continuity drove the method end to end.

Migrating 15 years of data to a cleaned model. 200,000 records with the inconsistencies fifteen years of use produce: duplicates, orphaned references, heterogeneous formats. Migration was as much about data quality as it was about transfer.

Transforming the organisation, not just the software. Taking a production team from 5 to 2 people via AI is a human problem before a technical one. The gradual approach and reskilling toward verification enabled a controlled transition.

Leading it all remotely. Team in Yaoundé, client in Montpellier: success in full remote relied on short and regular rituals, systematic demos and total transparency on progress.

What this project proves

• rebuilding a critical production system without interruption;
• recruiting and supervising a distributed team;
• leading through continuous client collaboration on an undocumented domain;
• change management: AI transformation of a production team;
• legacy modernisation: ASP/SQL Server → Next.js/Express;
• migrating 200,000 records with model cleanup;
• remote technical leadership over 3+ years of continuous collaboration.