Ounadjela Advisoryremiounadjela.com
Independent AI strategy & evaluation

Decide which AI pilots
to fund, fix, or stop.

I'm Remi Ounadjela. I work with leadership teams on the AI decisions that are expensive to get wrong: a safety review, a compliance report, a warehouse, a fleet. A good demo tells you very little there.

Book a 30-minute call See how the Sprint works

Start with a 30-minute conversation about one use case.

Former TikTok AI Strategy LeadPreviously Amazon and GoogleStanford instructor · Summer 2026 & Spring 2027Quoted in Fast Company

Call me when…

  • Every site or team has built its own pilot, and you need one way to compare them.
  • A pilot looks promising, but nobody has defined what result would justify scaling it.
  • The board wants the list: what gets funded this year, and on what evidence.
  • A vendor is proposing an AI project and you want an independent view before you commit.
01Working together

The starting point

AI Strategy Sprint

2–4 weeks Fixed scope

For leadership teams with more AI pilots than budget. In two to four weeks, you get a recommendation on each one and a plan for the next 90 days.

The standard plan

  1. Week 1

    Listen

    I talk to the people who own the candidate use cases and the budget, and collect what exists: pilots, vendor proposals, results so far.

  2. Week 2

    Decide

    One leadership workshop. We run each use case through eight questions, including value, error cost, reversibility, oversight, and evidence. You leave with a shortlist and a recommendation on each one: fund, fix, or stop.

    See the eight questions
  3. Weeks 3–4

    Plan

    I write the architecture and evaluation plan for the one or two use cases worth pursuing, with the criteria for deciding whether to proceed, and a 90-day plan your leaders and technical team can act on. We close with a readout to your sponsor.

    See a sample readout

This is the template. We adjust it to your decision and your calendar before we start.

What it takes from youA sponsor who owns the decision, about 45 minutes from each person I interview, and one half-day workshop with your leadership team.

Discuss a Sprint

We agree the decision, deliverables, and fee before starting.

02Proof

This isn’t theoretical for me.

Operating record

Built a function from zero
First Trust & Safety data science hire in EMEA at TikTok. I built the team across Dublin, San Francisco, and New York, and founded an academy that trained 600+ employees in AI and machine learning.
Wrote one standard for six systems
At TikTok, different teams had built six production LLM systems on 11 separate stacks. I wrote the plan to converge them and the standard for testing a system before it ships: human-labeled test sets and one evaluation method.
Shipped AI in operations
At Amazon, an ML model that optimized UK warehouse inventory placement and saved $4M a year.
“Remi is the leader I’d want deciding which AI work deserves investment.”
Nick JenkinsReported to me at TikTokFrom his LinkedIn recommendation

Advisory case

What to build first, and the evidence to fund it.

Trade finance compliance startup · 2025–2026

The founders knew trade finance. They had to decide what to build first, and they needed a technical plan they could take to engineers and investors.

The result: the founders used the plan as the technical foundation of their investor pitch.

“We knew trade finance. Remi turned it into a buildable AI product: the architecture, the evaluation plan, and a 12-week roadmap an engineering team could execute. When investors challenged the technical plan, his work was what we used to answer them.”
Founder, trade-finance compliance startup
More about this project

As sole technical advisor, I designed automated document examination around banking-rule checks, AI interpretation for ambiguous cases, and human review. The plan included citations, an audit trail, a security model, and cost estimates.

What I delivered

Architecture
Rules, AI interpretation, human review, and an audit trail.
Evaluation
A test design using the client's historical transactions.
Next steps
A 12-week engineering roadmap, hiring plan, and cost estimates.

A tool you can use

The AI Use-Case Triage Canvas

Eight questions to decide whether an AI use case should be piloted, scaled, paused, or rejected. With a worked example: Klarna’s customer-service assistant.

See the canvas and the worked example
03About
Remi Ounadjela
San Francisco Bay Area, 2026

I’m Remi Ounadjela.

Eight years across Amazon, Google, and TikTok, most recently leading AI strategy for TikTok’s global Trust & Safety organization.

Jan 2023 – Jun 2026

TikTok · AI Strategy & Adoption Lead

Nov 2020 – Jan 2023

TikTok · Senior Data Science Manager

I connect the leaders deciding whether to fund AI with the people who will build it. My work covers priorities, architecture, and evaluation. Your engineering team builds and operates the system.

AI Strategy. Stanford.

In summer 2026, I designed and taught a five-week course at Stanford for 45 industry practitioners. It returns in spring 2027 as “AI Strategy: How to Evaluate AI and Decide What to Do Next.”

Summer 2026 listing: AI and Accountability (TECH 75), Stanford Continuing Studies

Ounadjela Advisory is an independent practice, separate from Stanford.

Based in the San Francisco Bay Area. Available globally, in English and French.

Find where AI creates value. Prove it before you scale it.

04Get in touch

Bring me one use case.

Bring one AI use case and the decision you're trying to make. We'll use a 30-minute call to see whether I can help. Calls in English or French.

Book a 30-minute call

Choose a time in your time zone. We'll meet on Google Meet.