Data Quality Specialist

We're Ooak Data

Ooak Data turns company data into training data for AI agents.

Frontier labs can train models to reason. They cannot train them to work: navigating a real company's Slack threads, half-finished Notion docs, contradictory Jira tickets, and permission boundaries. That requires real enterprise data, and you cannot synthesize it. You have to source it.

We plug into enterprise tools, anonymize everything into a structurally identical digital twin, and generate reinforcement-learning environments with expert-level tasks calibrated against frontier models.

We are a Y Combinator company with 8 figures contracts signed with frontier AI labs, and we are scaling delivery aggressively over the next months.

🤝 Our values

• Trust, the foundation of every relationship, internal and external. We extend it by default and value the ownership that comes with it.
• Ambition, we commit, we move fast, with tenacity and efficiency.
• Collective, team first, low ego.
• Kindness, at the heart of every interaction.

The offer

📍 Location: Paris (city center), on-site with 1 to 2 days WFH

🕐 Start: as soon as you are available

💼 Contract: full-time (CDI)

💰 Compensation: fixed salary + equity (BSPCE)

✅ Your missions

As Data Quality Specialist, you are the last line between a dataset and a frontier AI lab. Nothing ships without passing your bar.

You sit inside the tech team, embedded with the engineers who build the ingestion and anonymization pipeline, and you work every day alongside the Ops team who run the deliveries. You report to Grégoire (COO & co-founder).

This is a hands-on, individual contributor role. You own the quality function itself, not a team.

• Define the standard. Acceptance criteria, checklists, sampling methods, error thresholds. Today they barely exist. You write them, and you make them stick.
• Audit what we ship. Processed and anonymized data, reviewed before delivery: PII leakage, replacement consistency, structural integrity, and whether the business value survived the anonymization.
• Automate your own job. Manual review does not scale. You write the detection scripts and the analyses yourself, and you work with the engineers to turn the ones that prove themselves into production checks, so the obvious errors never reach a human again.
• Measure and escalate. Quality metrics per dataset and per vendor, including the third parties we work with. You track them, you surface the problems, and you drive them to resolution.
• Translate client requirements into specs. Work with the tech and sales teams to turn what a lab actually wants into concrete, testable quality criteria.
• Close the loop with engineering. You are the person who tells the pipeline team what is broken upstream, with the evidence to back it. Quality problems get fixed at the source, not patched at delivery.

🔍 Who we're looking for

• 4+ years in data quality, data operations, or QA, ideally at a data labeling company, a data provider, or an AI company.
• Hands-on with large, messy datasets: processing them, auditing them, finding what is wrong with them. Python and SQL are required: you write your own scripts to interrogate a dataset, you do not wait for someone to pull the data for you.
• Comfortable in a tech team. You will sit with engineers, read their pipeline, and hold your own in a technical conversation. You are not expected to ship production code, but you are expected to be credible.
• A track record of building quality standards from scratch, not just applying someone else's checklist.
• Obsessive attention to detail. You enjoy finding the error everyone else missed, and you are not satisfied until you understand why it was there.
• Systems thinking. When you find one error, your instinct is to ask how many others like it exist and how to catch them automatically.
• AI-first. You use LLMs and AI tools as daily leverage to audit, automate, and move faster.
• Fluent English, mandatory (working language with our partners and the labs).

The extras that make the difference

• Familiarity with PII, de-identification, or privacy requirements.
• Exposure to annotation quality concepts: rubrics, inter-annotator agreement, guideline design.
• Understanding of how training data is actually used: SFT, RLHF, RL environments, evaluation.
• Experience working with external vendors or an annotation workforce.

🎁 Why join us

• A function to build, not to inherit. Quality at Ooak Data is yours to define from the ground up.
• Your work is visible. What you validate goes straight to the largest AI labs in the world.
• Entrepreneurial adventure: founding impact, live the early days of an ambitious company at the front line of the AI revolution.
• Backed by Y Combinator, join just as we accelerate.
• Offices in central Paris.
• Perks: Alan health insurance (mutuelle), 50% Navigo covered.

Interview Process

✏️ Recruitment process

We move fast.

• A 30 min intro call with Grégoire (co-founder)
• A 45 min technical deep dive with Thomas (CTO & co-founder): how you work with data, what you have built, how you think about detection
• A technical case: auditing a real anonymized dataset, debriefed with Grégoire and Thomas
• A final fit interview with 2 founders