AI Tribe - FinOps Analyst - Freelance

A) Context — building the AI Factory

AI Factory: an industrial organisation responsible for designing, producing and exposing the artificial intelligence capabilities made available to its entities across some thirty countries.

The AI Factory is building a catalogue of artificial intelligence products, platforms, frameworks and services, to be subscribed by Group entities through the marketplace. That catalogue rests on a simple principle: an offer is published only if it is genuinely consumable — documented, backed by a service level, a subscription journey and an established economic model.

The technical scope includes data science and generative AI platforms (Dataiku, Domino), a document processing chain (ABBYY), LLM as a Service and RAG as a Service offerings, conversational agents and a robotic process automation platform, all deployed on the BNP Paribas cloud. These offers consume vendor licences, cloud resources and inference services whose cost models differ profoundly.

It is precisely that economic model which remains to be built. None of the catalogue offers currently has a price, a billing unit or an established chargeback mechanism — and that absence conditions the publication of the entire catalogue. Furthermore, the relationship between the capacity provisioned to support these offers and their actual consumption is neither measured nor steered.

The assignment consists in establishing that model, instrumenting it and keeping it alive.

B) Required actions :

Build the pricing model for the catalogue offers, by family: products, platforms, frameworks and services.

Define, for each offer, the billing unit and the metering mechanism that makes it measurable.

Put in place the monitoring of actual consumption of the offers and of the technical resources supporting them.

Identify and drive the correction of gaps between provisioned capacity and observed consumption.

Produce the chargeback to consuming entities and guarantee its readability.

Feed the economic blocks of the catalogue entries and the AI Factory's financial governance bodies.

Scope boundary

The incumbent builds the economic model and guarantees its accuracy; pricing and commitment decisions rest with AI Factory management and Procurement.

Cost assumptions are established together with the technical teams operating the platforms, not settled in their place. Every figure produced must be explainable to the team concerned.

The incumbent does not manage vendor contracts, which remain with Procurement, but supplies the consumption and sizing evidence they require.

C) Main activities :

Building the pricing model

Establish the full cost of each offer: licences, infrastructure, inference consumption, operations and support.

Define the billing unit specific to each family — API call, document or page processed, token consumed, environment provisioned, expert day — and document the rationale for the choice.

Separate provisioning charges from consumption charges, and build the corresponding rate cards.

Set price review thresholds and the conditions triggering them.

Metering and instrumentation

Specify, with the architect and the platform teams, the per-consumer metering requirements: API gateway, quotas, logging, consumption attribution.

Verify metering reliability before any offer is published — unreliable metering renders the entire cost model unusable.

Build and maintain consumption dashboards by offer, by entity and by use case.

Consumption optimisation

Track the ratio between provisioned capacity and actual consumption across cloud environments, platform licences and inference services.

Identify overprovisioning, idle resources and poorly sized commitments, and issue costed recommendations with an expected saving.

Track token consumption and inference cost for LLM and RAG services, and propose optimisation levers: model selection, caching, quotas, instance sizing.

Prepare sizing decisions with the platform teams and Procurement.

Chargeback and governance

Produce the periodic chargeback to consuming entities, at a level of detail allowing each of them to understand and challenge its invoice.

Prepare the financial inputs for governance bodies: budget, consumption, variances, forecasts.

Feed the "economic model" and "measurement" blocks of the catalogue offer sheets.

Contribute to the value analysis of the offers, relating observed cost to actual usage.

D) Expected skills :

Assessed and weighted baseline for selection.

FinOps and cloud consumption management

Practice of FinOps principles: allocation, visibility, optimisation, forecasting.

Cloud cost analysis, in particular on Azure, and understanding of commitment models — reservations, reserved instances, savings plans.

Resource rightsizing, detection of underused capacity, management of quotas and budgets.

Understanding of cost structures specific to artificial intelligence: inference cost, token-based pricing, training cost, vector storage and query cost.

Offer construction and pricing

Building a full cost model and a rate card for internal IT services.

Defining billing units and implementing internal recharge mechanisms, both showback and effective chargeback.

Ability to articulate an economic model with a service catalogue and its commitment levels.

Tooling and data

Cloud cost management tooling, notably Azure Cost Management; familiarity with the FOCUS standard is an advantage.

Advanced spreadsheet command, SQL, and construction of decision-support dashboards (Power BI or equivalent).

Ability to work from heterogeneous, incomplete or uninstrumented consumption data, and to document the assumptions applied.

Working style and communication

Ability to put a cost question to a technical team without having the answer imposed — and without imposing it.

Traceability discipline: every figure produced must be justifiable, reproducible and defensible.

Ability to present an economic model both to executive management and to a technical team, in the terms of each.

Professional French and English.

E) Mandatory requirements :

Eliminatory prerequisites. An application not meeting all of these criteria will not be considered.

Minimum 5 years' experience in IT cost control, FinOps or economic steering of IT services, including at least 2 years on public cloud environments.

At least one experience building an internal recharge model for IT services, effectively implemented and applied, evidenced by a verifiable reference.

Demonstrated command of cloud cost analysis on Azure, or on another major platform with the ability to transpose immediately to IBM Cloud.

Advanced spreadsheet command and the ability to build a cost model auditable by a third party.

FinOps Certified Practitioner certification (FinOps Foundation) or a recognised equivalent.

Fluent French, written and spoken; professional English.

Availability to work on site at Rueil Malmaison 4 days per week..