About Disco

Disco powers a next-generation commerce media network that connects over 1,000 eCommerce advertisers to consumers on some of the world's most recognizable retail platforms, including Mindbody, Gopuff, Fabletics, and Knitwell Group (Chico's, Talbots, Lane Bryant). When a consumer books a yoga class or buys a new pair of jeans, Disco's recommendation engine surfaces relevant, brand-safe offers from other consumer brands, turning everyday transactions into meaningful moments of discovery.

Since launching in 2021, we've raised $26M from leading investors including Felicis Ventures, Bessemer Venture Partners, Shopify, Sugar Capital, RiverPark Ventures, and Indicator Ventures. Today, Disco's network spans 1,000+ advertisers and publishers, driving over $1B in GMV across 150M+ permissioned U.S. shopper profiles.

The Role

Disco is a commerce media company run like an AI company. We brought our bidding and ad-serving stack in house, and the machine learning inside it is the moat: how we price every impression, predict whether a user claims an offer, and route demand to supply to maximize yield and margin across three sides at once.

That machine learning is what you own.

Disco is a three-sided market. Publishers want more yield per load. Advertisers want more performance per dollar. Disco has to protect its margin in between. Every model decision trades one against the others: push CPMs up and CPCs drop, which hits margin. Optimize for publisher yield and advertiser performance suffers. Your job is to understand those trade-offs cold and turn them into better model decisioning, not better slides.

You'll partner with our two senior ML engineers to grow the precision and optimization of our bidding, our ad-server yield, our advertisers' performance, and Disco's margin. You won't implement the model, but you'll understand it well enough to push it: what data feeds it, where an eval is lying, which hypothesis to test next, and which decision makes the whole system better instead of one side better.

You'll do it the Disco way: with AI. You build with agents, you stand up eval loops and eval graphs instead of waiting on someone else to build your tooling, and you use AI to get a solution most of the way there yourself before pulling in an engineer. Exceptional AI tool use and ML thought processes are the same skill here, not two separate ones.

What You'll Own

Bidding, Yield & Performance Optimization

• Own the ML roadmap for our bidding and ad-serving engine: how we price, rank, and allocate every offer impression.
• Drive precision across all three sides at once: publisher yield (revenue per load), advertiser performance (CPA/ROAS), and Disco margin. Own the trade-off call when they conflict.
• Turn ambiguous optimization problems into hypotheses, evals, and shipped improvements. A CPA spike at Apple is your problem to chase to the source before it eats our ML engineers' time.

Model Decisioning & Trade-offs

• Own the decision logic that goes into the model: when we optimize for margin vs. publisher yield vs. advertiser performance, and how to make that call clearly.
• Think in systems, not single metrics. A CPM change ripples into CPC, which ripples into margin. You reason through the whole cascade before acting.
• Partner with the ML team on the full loop: hypothesis, model iteration, eval, A/B test, ship, monitor. You're a collaborative partner to the data science team, not a ticket writer throwing requests over the wall.
• Go deep on the underlying data. You know what a "load" actually is, how RPL is computed, and where the numbers can lie to you.

AI-Native Execution, Evals & Graphs

• Build with agents. Use AI to stand up the eval loops, dashboards, and eval graphs that track our optimization, and to automate the grunt work so the ML team stays on the hard problems.
• Own the eval loops and eval graphs that tell us whether a model change actually improved bidding, yield, or margin, and catch regressions before they ship.
• Productize what you build. A one-off script becomes a reusable eval, an owned graph, a versioned tool. You don't leave experiments as throwaway code.
• Be exceptional with AI tools, and keep current. The model and tooling landscape changes monthly; you track it so the team doesn't have to.

Customer-Facing Ownership

• Be the face of our optimization to customers. Walk into a tense room with a frustrated advertiser or publisher, own the problem, and hold it together with a clear plan.

Who You Are

• You're exceptional with AI tools and ML thought processes: you build with agents and reason through model trade-offs as one skill, not two.
• You have a real ML background. You've worked on or closely with ranking, bidding, prediction, or optimization models, and you can explain how they're built and where they break.
• You think in trade-offs and systems, not one metric in a vacuum. You see the whole marketplace: publisher, advertiser, and the margin between them.
• You're data-fluent. SQL, metrics, and model outputs are how you work. You can get into the data yourself.
• You form and ship hypotheses. You can generate a testable hypothesis on the spot and defend why it's right.
• You build eval loops and graphs, not just spec documents. You can show something you stood up yourself with AI.
• You're technically credible without being a full-time model builder. You can read code, query data, and reason about model trade-offs. You're not architecting the pipelines; you're the person who knows when and why to change them.
• Ad-tech background is a strong plus: agency, publisher yield team, ad-tech company, or third-party data tools. Not a hard requirement if you nail the ML, data, and customer fundamentals.
• You communicate across the team: a model trade-off to the CEO and a data question to an ML engineer in the same week.

Why Join Disco

At Disco, you’ll have the opportunity to work at the intersection of data, technology, and commerce, and have a direct impact on how our business and the broader ecosystem evolves.

• Unique Market Position: We're not competing on volume, we're competing on intelligence and yield optimization
• Data Advantage: 160M+ proprietary profiles give us differentiated insights and capabilities
• Innovation Runway: Ground-floor opportunity to shape our next generation of ad products
• Growth Stage: Opportunity to scale a proven business model into new dimensions
• Strategic Impact: Your work directly impacts both advertiser success and publisher economics

Perks & Benefits

We believe in taking care of our team, both inside and outside of work. When you join Disco, you’ll get more than great perks and benefits. You’ll be surrounded by great people, do meaningful work, and have the opportunity to build something that truly matters.

• Flexible PTO + 12 paid holidays + 3 company-wide DisConnect Days
• $250/month lifestyle stipend for the things that support your wellbeing
• $100/month Disco credit to shop from brands in our network
• $750/month childcare stipend for family care
• $132/month commuter benefit
• $500 home office setup reimbursement
• $20 daily Grubhub credit when working from our SF or NYC offices
• 401(k) to help you invest in your future
• Comprehensive insurance (medical, dental, vision & life)

This is a hybrid role based out of our NYC office, with an expectation of 2–3 days in the office each week. Compensation for this role is $190K–$220K OTE, plus meaningful equity.