Engine Builder (Operations Research)

ClosedWon Talent LLC; United States; On-site; USD 120,000 - 190,000; Full time

Job summary

ClosedWon Talent LLC is seeking an Optimization Engineer to own the core optimization engine, enabling production-grade planning across multiple clients. You will work on production deployments, ensuring traceability and auditable decisions while adapting models to changing requirements.

With a focus on Python and SQL, you will collaborate closely with engineering and data teams to deliver scalable AI-enabled optimization for our clients in the United States.

Qualifications

• Graduate degree in operations research, industrial engineering, or closely related field.
• 2+ years of real-world experience post-graduation.
• Built optimization models that shipped and were used.
• Strong Python and SQL skills.
• Comfort with AI as part of how you work.

Responsibilities

• Own the optimization engine end-to-end for production use.
• Ensure traceability and auditable decisions across planners.
• Tune and re-plan scenarios for changing conditions and clients.
• Collaborate with engineering and data teams to deploy models.

Skills

Optimization modeling

Python

SQL

Operations research

Problem solving

Education

Graduate degree in operations research or industrial engineering

Tools

SQL

Pandas

Job description

We’re ClosedWon Talent, a sales and go-to-market recruiting agency that specializes in working with fast-growing startups. If you’re exploring your next move, we might have just the opportunity for you, including this one.

About the Company

Most companies still plan their supply chains, meaning quantities, mix, allocations, and timing, in spreadsheets or legacy software. The spreadsheets break as soon as the business moves quickly. The legacy suites take a year to implement and frequently lose to the spreadsheets anyway.

Atomic is rebuilding the brain behind the American supply chain. By tackling operations optimization from first principles with an AI-native stack, they deliver unprecedented results at breakneck speed for their customers. Customers include DoorDash, HelloFresh, Archer Meat Snacks, Starface World, LMNT, OOFOS, Cora, and Chomps.

The founders lived this problem before they built the solution. Michael Rossiter and Neal Suidan built Tesla’s end-to-end supply chain orchestration system from scratch during the Model 3 ramp, after watching a company that size run on spreadsheets that could not keep up with how fast the business moved. Revenue has grown roughly 10x this year, and they just announced a $12.5M Series A, which is what’s funding this round of hiring.

About the Role

You’d own the optimization engine. It’s the core of the product, and it’s the piece every other part of the platform exists to serve.

Atomic runs over a million SKU by location by supplier combinations every day, and customers run 100+ scenarios daily against them. Every number the engine produces traces back to the logic that generated it at the unit level, because a planner reviews and approves it before anything moves. That traceability requirement rules out a lot of easy answers.

What the engine decides:

• Production planning. What gets produced, in what quantity, when, by which facility, and from what materials, against bills of material, minimum order quantities, run constraints, automation rules, and co-manufacturer capabilities
• Inventory allocation. Weekly positioning across a fulfillment network by SKU, by 3PL, by week, respecting channel splits, transit times, truck thresholds, reservation rules, and eligibility rules between specific SKUs and specific warehouses
• Purchasing. Generating purchase orders against supplier constraints, lead times, contract minimums, and cost, including grouping minimum order quantities across different SKUs that share components
• Scenario response. A tariff lands, a lead time slips, marketing pulls a launch forward, a new co-manufacturer comes online. The engine re-plans against the change, lets a planner compare any version to any other, and keeps a full audit trail of every override

The hard part is flexibility, not a single correct answer.

Every customer runs differently. One solves period by module, the next module by period. Some have multi-echelon constraints that others don’t. The engine has to be built so forward-deployed engineers can configure it per client without rewriting it each time. This is not a deterministic solver you tune once. It’s a platform that has to hold up across businesses that genuinely don’t resemble each other, and it’s about to get stretched further as Atomic moves into industrials and beyond CPG.

The engineering team is small and staying that way on purpose. There are no silos and no plans for any. You’d own this end to end, from the formulation through to what runs in production for customers.

Need-to-Have

• Graduate degree in operations research, industrial engineering, or a closely related field from a University like Georgia Tech, MIT, Michigan, Stanford, Berkeley, Columbia, Cornell, Princeton, Northwestern, Carnegie Mellon, Wisconsin, or Purdue
• 2+ years of real-world experience post-graduation
• You’ve built optimization models that shipped and got used, not just formulated and published
• Strong Python and SQL
• Comfort with AI as part of how you work

Nice-to-Have

• Supply chain optimization specifically. Experience at a planning vendor like Kinaxis or o9, or at a company that ran one, is a significant plus
• Consulting background, where you had to make models work for clients whose businesses differed from each other
• Multi-echelon inventory optimization, production scheduling, or network design
• Experience building configurable systems rather than one-off models