Florent
Salettes
Product Manager — Data & AI
Marseille, France
30-second read
Hands-on, from product strategy down to SQL.
Product Manager, Data & AI in B2B SaaS. Curious, autonomous, and comfortable building products as much as evolving them. My focus: data, API/MCP, and AI products. Oriented toward customer value, revenue, and growth. Hands-on, from discovery and product strategy down to SQL. At Opsealog: 350 vessels and 8 oil-major clients.
Four things to take away
Track record
Profile
Product Manager, Data & AI in B2B SaaS, at the crossroads of business and technology. Curious, autonomous, and comfortable building products as much as evolving them. My focus: data & AI products, APIs/MCP, and the governance that makes data usable by applications and agents.
Hands-on, from discovery and product strategy down to SQL. AI is as much part of how I work as of how I think about product: I use it, experiment with it, and build it in wherever it creates real value for customers.
My goal: build products that customers adopt and that generate revenue and growth for the business. English C1.
Skills
Product
Product vision and strategy, product discovery, customer interviews, Voice of Customer, opportunity qualification, prototyping and experimentation, roadmap, value-based prioritization, business cases and ROI, build/buy trade-offs, pricing and packaging, go-to-market, sales enablement, end-to-end delivery, Agile Scrum.
Data and AI
Data product management, data platform, medallion architecture, information architecture, data governance, data contracts, data quality, semantic layer, data sharing and monetization, API product design, data readiness for AI, AI product governance and evaluation (shadow testing, human-in-the-loop), AI agents, MCP.
Tools
French native, English C1.
Experience
Opsealog
B2B SaaS, offshore maritime performance — Marseille | Reporting to the CPTO | Working with a 5-person data squad | 350 vessels, 8 clients (oil majors)
- Product discovery and prioritization — Designed and rolled out the framework feeding the roadmap: 7 collection channels, around 200 customer and business signals broken down atomically, grouped by convergence into weighted insights, then converted into around twenty qualified opportunities for arbitration in the Offer Committee. Became the single entry point to the roadmap.
- API-first data product and monetization — Defined, launched and commercialized a data product giving access to data from 150 vessels through 5 endpoints exposed via API and MCP, for consumption by applications, client teams and AI agents. Structured the subscription model, governance, interface contracts, versioning, access terms and onboarding. Sold as an upsell to two clients and used in production.
- Self-service analytics AI agent — Owned an agent that answers natural-language questions and returns text, tables or charts on operational data. Defined capabilities and priority use cases, ran business validation, tracked product metrics: 90% qualitative response rate, 2-3 minute response time, classified errors, monitored cost. Deployed to production, currently in an internal testing phase: shadow testing preserving the human baseline, on a cross-functional panel of 10 users, ahead of opening to clients.
- Data governance and semantic layer — Used the exit from legacy — migration to dbt, medallion architecture — to define the data governance model with business teams and introduce a common semantic layer. Functional owner of inbound and outbound data contracts, business definitions, their lifecycle and quality standards. This layer will feed the MCP exposure and the analytics agent.
- Event-driven decision engine — Designed a configurable engine for qualification, scoring and action triggering, handed to business teams to reduce internal inefficiency: automated contacts, follow-ups and escalations. Prototyped a first version alone, tested in real conditions across the whole fleet, before it was industrialized by the technical team and extended to other use cases.
- From lookup tables to a predictive model — Defined the PRD and target architecture for a machine learning model intended to replace manually maintained, non-scalable lookup tables. Used more than ten years of data to build the dataset and select variables, set a 95% reliability target. Project approved in principle by leadership, subject to budget arbitration before launch.
- Partner IoT offer — Built the value proposition around integrating data from onboard sensors, in partnership with a manufacturer: manufacturer evaluation, defined the integration model into the value chain, designed the offer with the business team. Launched a PoC covering 25 vessels.
- Analytics delivery — Built and maintained datasets and analytics products used in self-service by internal teams and clients. Ran the analytics roadmap and delivery in Agile Scrum. Managed the data budget and the commercial relationship with data suppliers.
Teamwill Consulting
IT consulting firm
- Digital Project Manager — Ma French Bank assignment (04/2019 – 10/2020): launched the revolving credit offer on the business side, from requirements to deliverable sign-off. Subscription and drawdown journeys, functional specifications, user acceptance testing.
- Product Owner — GE Money Bank assignment (11/2017 – 01/2019): ran the roadmap and product backlog in Scrum for a securitization application: new features, application diversification, new operation types.
- Functional Project Manager — GE Money Bank assignment (12/2015 – 11/2017): took a securitization application live: full functional design (workshops, data models, interfaces), specifications, test management through go-live. Fixed-price waterfall project.
- Business Analyst — GE Money Bank assignment (01/2015 – 08/2015): formalized business needs, the functional solution and inter-application data flows during a pre-sales phase.
Side projects
Product Lab — AI-assisted prototyping harness — Pipeline from Markdown spec to deployed prototype: breakdown into development issues, published to GitHub, CI/CD and deployment on a Docker image. Turns an idea into a working prototype without pulling in a technical team.
Product opportunity qualification engine — Structured multi-source extraction, AI-assisted scoring, explicit human validation before any write. Stack: Claude Code, MCP servers, Groq API.
Hermes — self-hosted personal assistant — Assistant hosted on a Raspberry Pi, interface deployed on Vercel behind Cloudflare.
Education
Generalist Engineering degree (MSc) — Arts et Métiers ParisTech (ENSAM), Paris
References available upon request