Director of AI Data Solutions

Pearl is AI for professional services at global scale—combining advanced AI with verified human expertise to deliver help that’s accurate, accountable, and fast. Since 2003, our network has connected millions of customers with licensed professionals across 196 countries, making real expertise available anytime, anywhere.

• Courageous: We take risks and challenge the status quo
• Innovative: We're constantly learning, creating, and adapting
• Lean: We focus on customers, using lean testing to learn how to serve them best
• Humble: Past success is not a guarantee of future success

Pearl (the parent company of Just Answer) is building a category-defining enterprise business at the intersection of AI and professional human expertise. One of our most important bets is AI Human Data Services:
Expert data products that accelerate frontier model development - data that goes beyond generic labeling and commodity reinforcement learning through human feedback (RLHF) into high-signal, professional-grade work that materially improves model behavior.

AI labs are moving from models that “know things” to models that “do things.” That shift demands training and evaluation data that is harder, longer-horizon, and more domain-specific:
PhD-level human evaluations, expert preference selection, verifiers, rubrics, agentic task environments, and multimodal workflows. Pearl is uniquely positioned to win here because we have a large, liquid network of licensed professionals (>20,000 professionals), an operating platform built for high-trust expert work over the last 20 years, and multimodal production capabilities across chat, voice, and computer-use style tasks.

We need a AI data service entrepreneur/intrapreneur to discover where we can have maximum product-market fit and then build on early AI Data Services traction to build a durable product line: a repeatable set of data offerings, a world-class quality and evaluation system, and a delivery engine that frontier labs trust with their most important training and evaluation problems. This is not a quota-carrying sales role.

It’s a product-and-operations leadership role with deep customer intimacy - owning the end-to-end data product lifecycle from discovery to specification to production to quality to expansion.

We believe the ideal leader is someone who has either:

• Owned procurement / partnerships for custom AI training or evaluation data inside a major AI lab (or equivalent model-building org), or
• Successfully brought to market high-end custom data programs at a leading data provider (e.g., Scale, Appen, Handshake, or similar), with firsthand experience of what differentiates “elite” data from commodity output

Build Pearl’s AI Data Services business into the trusted, go-to partner for frontier labs and AI-native companies that need expert-grade data at speed and scale - without sacrificing rigor, confidentiality, or quality.

• Define and productize our AI Data Services portfolio (what we sell, how it’s packaged, how it’s priced, what “done” means).
• Build the delivery engine: staffing, workflows, tooling, QA loops, and automation that make elite data repeatable.
• Establish Pearl’s quality bar and measurement system: rubrics, verifiers, audits, gold sets, calibration, and continuous performance scoring for experts and outputs.
• Own customer outcomes: translate ambiguous research needs into precise task designs and production-ready specs that reliably improve models.
• Build data rights and licensing strategy in partnership with Legal: ensure we can responsibly reuse and monetize data across customers when appropriate, while honoring confidentiality, privacy, and contractual commitments.
• Create Pearl’s external credibility in expert data: benchmarks, evaluation frameworks, and clearly articulated standards.

This role spans multiple data modalities and program types, including:

• Expert Supervised Fine Tuning (SFT) and instruction data for specialized domains.
• Preference data / ranking / pairwise comparisons with high-quality rubrics.
• RLHF-style programs: verifiers, critiques, and structured…