Data Product Operations Lead | Audio AI Research Company | San Francisco | Up to US$180K + Equity

"Sound is the vocabulary of nature." — Pierre Schaeffer

The next generation of AI will listen before it reads. Somebody has to build the data it listens to. Our client, a venture-backed audio AI research company in San Francisco, is hiring a Data Product Operations Lead, in plain terms a Data Operations Manager for audio, to run the voice and speech datasets that frontier AI labs train on. On-site, up to $180K base plus competitive equity, visa transfers welcome.

Three things you must bring

Call it Data Product Operations Lead, Operations Lead for AI Training Data or Program Manager for Data Collection; the job is the same. Before you read another word, check yourself against these. All three are required, and the hiring manager screens on them first.

You have owned something from prototype to production. Not contributed to it. Owned it. You were there when it was a spreadsheet and three people, and you were still there when it ran at volume with metrics, QA gates and a delivery schedule. A marketplace launch, a data pipeline, an ops process built from nothing, a consulting pilot you took all the way through implementation.

You have 1–6 years in a high-intensity operator seat. Top-tier management consulting, strategy & operations at a fast-moving marketplace or tech company, or product operations between product, engineering and the customer. Up to eight years considered for the right person.

You have led cross-functional work across ops, product and engineering, and you have real product sense. You have driven workstreams where nobody reported to you, and you are comfortable being the non-engineer in the room who still makes the call.

If that is you, keep reading.

The company

Every voice assistant, real-time translator and call-centre agent is only as good as the speech it was trained on. Public audio is thin, noisy and legally murky, so purpose-built, consented, precisely specified audio data is the bottleneck for the whole category. Our client exists to remove it.

They are an audio AI research company, bringing the same rigour to building datasets that the best labs bring to building models. They design the spec, recruit and brief the voice contributors, run the collection, grade the quality and ship the dataset. Data is the product.

It is working. The founders came out of one of the best-known human-data companies in the world and have run this playbook at scale before. The company already counts most of the big consumer-tech names and the leading AI labs as customers, has closed a Series B from tier-1 investors with tens of millions raised, and has grown to a team of between 50 and 100 in one San Francisco office. Demand from the labs is outrunning the operations team, which is why this role is open.

What you will actually do

Think of yourself as the Data Operations Manager for a factory whose product is sound. You own audio and voice data projects from the first customer conversation to final delivery. A lab needs, say, 2,000 hours of natural conversation across a dozen accents with specific noise conditions and emotion labels. You turn that into a spec, work out the contributor pool, tooling, QA and budget, line up staffing, set the timeline and ship it on spec and on time. That is more than half the job. The rest:

Make the data better. Work directly with the voice-contributor community and partner with engineering and ML so quality climbs project over project.

Take pipelines from 0→1 to 1→N. New data types start scrappy. You prove the prototype, then turn it into a repeatable production line with throughput targets and cost controls.

Run the numbers. Hours per day, acceptance and rework rates, cost per accepted hour. SQL is how you get at them; if you do not write it yet, you will learn fast.

Face the researchers. Around a tenth of your time is spent with ML researchers at frontier labs on deliveries, trade-offs and quality sign-off. They are precise and quick to spot hand-waving.

There is no playbook to inherit. You write it.

Who thrives here

If your current title is Strategy & Operations Lead, Program Manager or Data Operations Lead, you are probably in the right place. The hiring manager's two words are "scrappy" and "analytical": builders who will listen to sample audio at 11pm one day and present a cost-per-hour analysis to a lab's research lead the next. Backgrounds that map cleanly:

Human data and AI training data. Scale AI, Surge AI, Mercor, Labelbox, Appen, Turing, Invisible, Toloka, Prolific and their peers.

Top-tier consulting. McKinsey, BCG, Bain and the tier just behind them, ideally with a hands-on implementation phase on your CV.

Marketplace and growth-stage strategy & ops. DoorDash, Uber, Lyft, Instacart, Airbnb, Faire, Rippling, Deel, Flexport: people who launched and scaled supply programmes on metrics.

Early-stage operators. The person who built the ops function at a seed or Series A company from nothing.

Product ops and TPM. Product operations at a top software company, or a TPM who has run cross-functional launches with engineers.

Audio and speech. Deepgram, AssemblyAI, ElevenLabs, Speechmatics, Rev, or speech teams inside the big platforms. Rare, and a real edge.

Nice-to-haves, and what this role is not

Nice-to-haves: customer-facing experience with technical stakeholders, a bachelor's degree from a top university in any field, working SQL, prior human-data or AI training-data experience, and any audio, voice or speech domain knowledge.

One thing rules you out: low ownership. If your instinct is to delegate the unglamorous work, this is the wrong role.

And what it is not: an ML or data science job, a data analyst job, a PMO job, or a classic software PM job. It is an Operations Lead role for AI training data: the product is a dataset and you own the outcome. It is also not remote, not hybrid and not a senior leadership hire.

The package, location and visa

Base salary: up to $180,000 USD, depending on experience.

Equity: competitive early-stage equity, a few months past a Series B.

Location: on-site in the company's San Francisco office, five days a week. No remote, no hybrid, and no relocation support, so you need to be in the Bay Area or able to move at your own cost before start.

Visa: US citizens, green card holders, OPT holders and existing H-1B holders (transfers supported) are all welcome. The company cannot sponsor a new visa from scratch.

Ready to be heard?

Text taught machines to read. Voice is about to teach them to listen, and the models that win will be the ones trained on the richest, cleanest, most human sound on earth. The Data Operations Manager who builds that data does not need a PhD. They need to own the whole thing, from a rough first recording to thousands of hours shipped to a frontier lab.

If that is you, send your CV and a short note telling us about one project or pipeline you took from prototype to scale, with the numbers that prove it. Add your current location and whether you can be on-site in San Francisco, your US work authorisation and your base salary expectation. Hit the three must-haves head-on and you will be in front of the hiring manager within days. The next wave of AI is listening. Make sure it hears you.