PRINCIPAL FORWARD-DEPLOYED ENGINEER

STATION F; Paris; Hybride; EUR 83 000 - 152 000; Plein temps

Résumé du poste

STATION F is seeking a freelance ML/AI platform engineer in Paris to own the end-to-end delivery of a reasoning layer for multi-site operators. You will design an operator workbench, define evaluation harnesses, and deploy a robust, auditable system inside a client's environment.

You will collaborate with front-end engineers, enforce strict API contracts, and balance standardization with bespoke deployment patterns to deliver secure, scalable solutions for real-world clients.

Qualifications

• Experience shipping a product where a model is part of a multi-component stack.
• Experience with evaluation harness, regression suites, drift monitoring.
• Ability to work in English; French a plus.

Responsabilités

• Design and implement an operator workbench used by non-AI users.
• Define the evaluation harness, retrieval, scoring and gating.
• Architect and maintain API contracts separating backend, agents, and frontend.
• Enforce security, identity, data boundaries, and least privilege in client tenants.
• Lead standardization while enabling reusable patterns for new clients.
• Collaborate with frontend engineers to deliver end-to-end solutions.

Connaissances

Evaluation-first

Ownership

Frontend

Architecture

Security

Outils

TypeScript

Frontend framework

Description du poste

Job Description

Fieldreason builds the reasoning layer over systems of record for multi-site operators: companies running dozens or hundreds of plants, sites, vessels or clinics far from head office. The ERP knows what was ordered. Nobody's system knows what should be ordered, what will expire first, what can legally cross which border, or which alternative is acceptable when the exact item is gone. Everyone can call the same models now, so nothing is won at the model layer: the work, and the difference, is in evaluation, orchestration and the interface an operations team actually uses. We put that in software, ship it live inside the client's environment in weeks, and stay to run it. Our first client runs 260 sites.

Fieldreason builds the reasoning layer over systems of record for multi-site operators: companies running dozens or hundreds of plants, sites, vessels or clinics far from head office. The ERP knows what was ordered. Nobody's system knows what should be ordered, what will expire first, what can legally cross which border, or which alternative is acceptable when the exact item is gone. Everyone can call the same models now, so nothing is won at the model layer: the work, and the difference, is in evaluation, orchestration and the interface an operations team actually uses. We put that in software, ship it live inside the client's environment in weeks, and stay to run it. Our first client runs 260 sites.

What You Will Do

• Evaluation first. Build the harness before the agent: golden sets from real cases, regression on every model, prompt or provider change, drift monitoring in production, every exception traceable to its inputs. Public benchmarks decide nothing here; a model that wins on MMLU under one prompting technique loses under another.
• Orchestrate. The chain is not one call: intent classification, retrieval from the client's own records, resolution, scoring of the proposal, human gate on exceptions. You decide what each step is made of and where a plain rule beats a model.
• Build the surface. The first deliverable is an operator workbench used every morning by people who did not ask for AI. You own it end to end, or you work with a front-end engineer and hold the standard. A reasoning layer nobody trusts enough to click is worth nothing.
• Deploy adversarially. Identity, data boundaries and least privilege inside the client tenant, prompt-injection and tool-abuse defences, an audit trail that survives an auditor.
• Keep backend, agents and front end separated by explicit API contracts, so the reasoning layer can be swapped without touching the interface.
• Make disciplined standardization-versus-customization decisions instead of defaulting to bespoke work; turn messy deployment problems into reusable patterns for the next client.

Preferred Experience

• You have shipped a product where a model was one component among databases, queues, contracts and a screen, and real users depended on the result.
• You work evaluation-first: criteria defined before the build, golden sets from real cases, regression on every model, prompt or provider change, drift monitoring in production.
• You have built the layer that watches the system, not only the system: evaluation harness, regression suite, cost and latency per call, drift and exception rates, ideally across more than one build.
• You can build the surface: TypeScript and a modern front-end framework. The first deliverable is an operator workbench used every morning by people who did not ask for AI.
• You have delivered inside a client's cloud tenant, ideally Microsoft: Fabric and Foundry, Entra ID, infrastructure as code. The client's CIO probes this first.
• You are sound on state that must be right: inventory truth, expiry, country import rules. Relational where correctness matters, document stores where honest, graph for catalogue equivalence, vector indexes for what they are and are not.
• You can explain what temperature, top-p, top-k and the seed actually do, why fixing them still does not buy determinism, and what you did about it in a real system.
• You have enforced output formats where it mattered, and have an opinion on constrained decoding versus finetuning versus a validation pass.
• Senior enough to own the build end to end: you write the specification before the code and can argue calmly with a compliance officer.
• English. French a plus.

Recruitment Process

• Thirty minutes with Antoine, Managing Partner.
• A sixty-minute technical screen with a senior forward-deployed engineer: a synthetic exercise, deploying a resolver into a client Azure tenant (identity, data boundary, model access, infrastructure as code, where the confirmation gate sits), plus two questions we care about, how you would know the system got better after a model change, and how you would stop an agent from treating its own previous output as fact. We grade the reasoning, not the syntax.
• One reference who vouches for one thing you shipped inside a client tenant.
• Terms and start.

Additional Information

• Contract Type: Freelance
• Location: Paris
• Occasional remote authorized