AI Engineer, Optimisation & Intelligence | San Francisco (Onsite) | $200K–$300K base

Inventure; San Francisco (CA); On-site; USD 200,000 - 300,000; Full time

Job summary

Inventure in San Francisco is seeking an ML Engineer to own the intelligence layer of ad bidding and budget optimization, with hands-on responsibility for decisions that directly impact live campaigns.

The role requires deep production experience with closed-loop optimization, RL or bandits, and a track record of delivering real ROI from live spend. This is a full-on onsite position in SF, visa sponsorship available.

Responsibilities

• Build the decision engine: bid changes, budget reallocation, pause and boost calls, and postback optimization.
• Design the learning loop so outcomes continually improve campaign strategy.
• Write optimization policies from simple rules to reinforcement learning.
• Coordinate multiple agents reasoning over campaign context and their outputs.
• Own the number: accountable for spend returns and decision quality.

Job description

ML Engineer, Ad Bidding & Budget Optimization | San Francisco (5 days onsite) | $200K-$300K base

We've partnered with one of the fastest-growing AI startups in San Francisco, building an autonomous growth platform that runs paid customer acquisition end to end. Their agents replace the traditional media-buying stack, letting companies run paid acquisition across the major ad platforms without human media buyers. The traction is exceptional and remarkably capital-efficient: around $30M in revenue growing roughly 30% month on month, all on a lean $10M raised, with live spend running across Meta, Google, Tik Tok and Snapchat for customers in mobile, gaming, AI and tech. They are now raising a Series A on the back of it. It is a small, flat, high-talent team winning on results rather than noise.

Who this is for

One of these three backgrounds fits you:

• An ML engineer who has owned a production closed-loop optimization system, where logged outcomes changed what the system decided next.
• A quant, or an RL and bandits specialist, moving into applied ads optimization.
• An ads-optimization engineer who has worked directly on live spend: bidding, pacing, budget allocation, postback optimization.

If you know what a postback is and you have watched a pacing decision play out against a ROAS target on real money, keep reading. If your production experience is RAG pipelines and agent frameworks without the spend behind them, this one is not for you.

This is a five-day-a-week onsite role in San Francisco. No hybrid, no remote, no exceptions.

The role

This is a rare opportunity to own the intelligence layer of the product, the part that actually decides what happens to a campaign. A dedicated platform team builds the simulator and the tooling; you own the decisions, the policy and the learning loop. Because the system manages real ad spend, every decision you design is graded against returns within days, not quarters. You'll report directly to a hands-on, technical CTO who is in the code daily, work with genuine autonomy, and get the intellectual pull of quant-style optimization applied to live ads rather than a trading book. If you're the kind of engineer already tinkering with something new this week, you'll feel at home.

You will:

• Build the decision engine: recommendation and scoring systems for bid changes, budget reallocation, pause and boost calls, and postback optimization.
• Design the learning loop that defines how the system learns from outcomes, so campaign strategy compounds over time rather than resetting.
• Write the optimization policies over noisy live data, from simple rule-based strategies through post-training and reinforcement learning.
• Orchestrate multiple agents reasoning over campaign context, design how they work together, and evaluate what they produce.
• Own the number: you are accountable for decision quality and what the spend returns, not for platform plumbing or simulator infrastructure.

About you

• You have shipped a closed-loop decision or optimization system in production, and you can point to the method you used and the metric it moved.
• You are hands-on with optimization over noisy, live data using techniques such as reinforcement learning, bandits, post-training or rule-based strategy.
• You have direct experience with ads mechanics such as bidding, budget and ROAS, rather than only search or ranking systems.
• You are comfortable designing and orchestrating multiple agents that reason over context, and evaluating their output.
• You have a recent production track record, ideally with systems shipped in the last year and a strong GitHub or production history working with live data.
• You thrive on autonomy and ownership, and you have a genuine bias toward building.
• You are based in San Francisco or willing to relocate, and happy to be in the office five days a week. Visa sponsorship is available, including OPT and H1B transfers.