Human Data Manager

About BeatpulseLabs

Beatpulse Labs builds high-quality human data for some of the world's most demanding AI applications. We run large-scale data programs across video, voice, music and other multimodal datasets, combining global contributor networks, domain expertise and rigorous quality systems to deliver training and evaluation data at scale.

As we grow, we're looking for exceptional operators who've built or run complex human-data programs and want end-to-end ownership of a major vertical.

The Role

• You'll own the execution and scaling of human-data programs, from client brief through workflow design, contributor recruitment, production, QA and delivery. This is hands-on and high-ownership: you won't just manage projects, you'll build the operating systems that let hundreds or thousands of contributors and experts consistently produce quality data at scale.
• We value deep expertise in one modality over shallow experience across many. Tell us where you're strongest: video, voice/audio, music, or another human-data vertical.

What You'll Own

• End-to-end operations. Turn complex or ambiguous client requirements into scalable production workflows. Design annotation, collection, evaluation and QA processes, and build playbooks that hold up as volumes climb.
• Workforce at scale. Recruit, qualify and structure distributed networks of annotators, reviewers and experts. Forecast capacity against deadlines and keep improving quality, reliability and throughput.
• Team management at scale. Manage teams of thousands or tens of thousands of people, with clear structures, accountability and operating rhythms that maintain quality and throughput at scale.
• Quality & delivery. Own the quality of the data that reaches the client. Set acceptance criteria, build QA systems, and catch systematic failures before they hit delivery.
• Client delivery. Turn evolving client requirements into executable programs alongside commercial teams. Communicate risks and trade-offs clearly, and own delivery against quality, timeline and commercial expectations.
• Metrics & unit economics. Track throughput, acceptance rates, productivity, turnaround, cost per accepted unit and margin, and use them to clear bottlenecks and build sustainable economics.
• Tooling & automation. Partner with product and engineering to sharpen contributor tooling, automate the repetitive, and design systems for far larger volumes and populations.

What We're Looking For

• 5+ years in human data, AI training data, annotation, AI evaluation, marketplace ops or a related field, or equivalent experience running complex programs at scale.
• Direct ownership of high-volume human-data programs, ideally spanning hundreds or thousands of distributed contributors or experts.
• Experience designing annotation, evaluation, data collection or QA workflows.
• A strong grasp of the trade-offs between quality, throughput, capacity, speed and cost.
• Strong analytical and operational problem-solving, comfortable in fast-moving environments and moving between strategy and hands-on execution without a large team beneath you.
• Strong client-facing and communication skills.
• Ownership and a hustler mentality. When something breaks you find the root cause and fix it; you're scrappy, biased toward action, and don't wait for permission or perfect information.

Experience at a scaled AI-data company (Surge AI, Mercor, Scale AI, Handshake AI, micro1 or equivalent) is strongly preferred, especially launching programs from scratch, building qualification and QA systems, improving acceptance rates or cutting cost per accepted unit, working directly with AI labs, or owning project P&L.

RLHF, SFT, model evaluations, preference data and other post-training experience is valuable but not required.

Modalities We're Hiring Across

• Video: annotation, collection, scene description, visual understanding, action recognition, multimodal video.
• Voice / Audio: speech collection, transcription, recording, audio annotation, speech evaluation, conversational data.
• Music: music annotation, expert evaluation, composition-related datasets, music understanding, other specialist music data.
• Other human data: exceptional operators from other verticals who've built and scaled complex AI training or evaluation programs.

What Success Looks Like

Within your first few months, you can take a new client requirement and independently build the operating system to deliver it: the right contributors, task and workflow design, QA, capacity, metrics and economics. You'll know where an operation breaks at 100, 1,000 and 10,000 contributors, and how to get ahead of it. Ultimately, we want people who can own a major human-data vertical at Beatpulse and build it into a world-class operation.

Who Will Thrive Here

You'll enjoy Beatpulse if you prefer building systems over inheriting them, make good calls with incomplete information, and move fast without dropping quality. You're highly analytical but just as comfortable with people, you care about both client outcomes and project economics, and you want to build the human-data infrastructure behind the next generation of AI.