Tami :

Find out how we took over and evolved Tami's MVP, an HR platform where AI conducts and transcribes job interviews, to give it the solid foundations it needed to grow.

HR Tech
AI
MVP takeover
Web app

Client: Tami
Sector: HR Tech / AI recruitment
Deliverables: MVP takeover and evolution, AI web app
Expertise involved: Code takeover, web development, AI, UX/UI
Tami's product: mytami.fr

Tami is an HR solution that runs job interviews with AI: an artificial intelligence actually conducts the interview out loud, then transcribes and analyses the conversation to save recruiters precious time. A premium positioning, aimed at demanding recruiters and major brands (Accor, Carrefour, the YOU hotel…).

A first MVP had already validated the concept and won over its first users. But as is often the case, the solution that made it possible to move fast at the start wasn't designed to last: it saved time at launch, then began to show its limits.

The challenge for Tami? Past validation, the product had to become sturdier to welcome more customers and keep evolving, without breaking with every new feature.

To cross that threshold, they needed a team able to take over what existed, make it reliable, then evolve it at the pace of their growth.

The real challenge: take Tami from MVP to robust product, without starting from scratch or losing the lead already gained.

We had to take over a product we hadn't built, understand how it worked, and spot what was solid and what was fragile. Then we had to lay healthy foundations while keeping the product running for its users.

And because this is AI recruiting, we had to go all out on the sensitive topics: reliability of the analyses, confidentiality of candidate data, GDPR and the AI Act. All requirements to meet while the product kept evolving.

1.Takeover and audit of the existing product
We started by analysing the MVP as it was: understanding how it was built, what held up and what was likely to break as the number of users grew.
2.Securing the foundations
Before adding a single feature, we cleaned up and secured the base: a clean, tested and reversible foundation, so the product could evolve without every change becoming a gamble.
3.Evolving to support growth
Once the foundations were sound, we resumed development: new features and a controlled scale-up. Tami keeps the same product, without the fragility of its early days.

Weekly sync meetings

Every week, a progress meeting let us validate what had been done, adjust priorities and prepare the next steps.

A shared audit from day one

Before touching the product, we laid out what existed, separated what was solid from what needed reworking, and set the priorities together.

Continuous test environments

As development progressed, we gave the client pre-production environments (Staging). As soon as a section was ready, the client could use it in real conditions and give us feedback so we could iterate quickly.

The client keeps full control at every step and sees their product evolve in real time: no black box, no bad surprises at delivery.

To make the product reliable and evolve it without sacrificing performance, we worked with the following technologies:

Front-endReact 18, TypeScript, Vite, Zustand, React Router, styled-components, socket.io-client
Back-endJava 21, Spring Boot 3.5 (Web, WebFlux, WebSocket), Spring Security + JWT, OpenAPI; two Python microservices over gRPC (speech detection and end-of-turn detection)
DatabasePostgreSQL (Spring Data JPA), Redis and Caffeine for caching
AI and videoOpenAI and Groq (LLMs), Deepgram (transcription), Cartesia (speech synthesis), Daily.co (video)
Hosting and monitoringDocker Compose, Prometheus / Micrometer, Grafana