Jim Rebesco and the problem with tomorrow

A hurricane arrives, refineries shut down, and an algorithm carries on with its calculations. It has plenty of data. It may also have a thoroughly obsolete picture of the world. Jim Rebesco likes this example because the difficulty is easy to see: yesterday’s useful forecast can become today’s expensive mistake without a single line of code breaking.

For an airline buying fuel, or an energy producer trying to anticipate prices, the change matters. Someone watching the news might recognize the problem before the statistical machinery does. Rebesco’s question is what happens next. How quickly can people gather fresh information, prepare it, retrain the model, and return something useful to service?

That question has followed him from financial trading into Striveworks, the Austin company he co-founded and leads. It gives his career a curious continuity. The settings change considerably. The trouble remains recognizable. A model has learned something about the world, and the world has decided to move on.

The hurricane is an example he has used in an interview, rather than a scene from a particular trading day. Its usefulness lies in how little explanation it needs. Weather has a talent for editing assumptions.

A scientist learns the market’s timetable

Rebesco’s education began with physics at the California Institute of Technology, where he earned his bachelor’s degree in 2004. He completed a doctorate in computational neuroscience at Northwestern University in 2009. The route gave him a scientific foundation for work that would become increasingly concerned with the practical behavior of algorithms.

At Virtu Financial, an electronic market-making firm, he became a partner and led trading and data science teams. His professional biography credits him with helping build the company’s analytical capabilities and contributing to its 2015 initial public offering. It was a setting where calculations had to operate inside a competitive market, with consequences arriving on the market’s schedule.

In explaining the connection to his later work, Rebesco has emphasized the effort needed to keep models online, functioning, and adapting. That is a different kind of achievement from producing a successful result once. An operating system needs attention through changes in its environment. The work continues after the person who built it has finished explaining why it should work.

Financial markets provided a demanding education in that distinction. Rebesco would eventually carry the problem into national security, where the surrounding conditions can be equally unwilling to cooperate.

The expert beside the algorithm

Rebesco gives the human specialist a substantial role in this story. When circumstances shift, a data scientist may need someone who understands the particular market or operating environment to decide whether the new information resembles the data used for training. Technical fluency alone does not settle the question.

His interest is in making the steps around that judgment connect more easily. Training, deployment, monitoring, and retraining should require less manual coordination. He has described the possibility of turning an end user’s acceptance or rejection of a suggestion into information that helps improve the system. The user is already reacting. The challenge is to make that reaction useful without requiring a separate administrative ceremony.

“I’m a data scientist at heart.”Jim Rebesco

He has also questioned how much attention developers devote to model architecture compared with the data used to train it. An extra hour spent improving the data may accomplish more than an extra hour spent adjusting the model. In his account, personalization adds another complication: the model suited to one person may differ from the model suited to another.

That makes validation and auditability practical concerns. What produced this answer? Which data informed it? How should an error change our confidence? Statistical models require ways to investigate their behavior. A polished answer supplies very little of that history by itself.

There is an appealing modesty in the ambition to make this machinery ordinary. In his 2022 discussion of the industry’s open problems, Rebesco wanted the supporting stack to become easier to use and less intrusive. Successful infrastructure often receives its compliment indirectly: people stop having to discuss it every morning.

A business with paying critics

Striveworks was founded in Austin in 2018 by Rebesco, Craig Desjardins, Eric Korman, and Anthony Manganiello. Its product, Chariot, addressed the processes around bringing machine learning into use and keeping it useful. Rebesco’s ambition extended to the experience of the people doing the work: maintaining analytical performance should happen quickly, behind the scenes, and with as little disruption as possible.

In June 2023, the company announced a $33 million first institutional round led by Centana Growth Partners. Institutional capital arrived after the business had already spent years developing its products. The sequence matters because it places customers earlier in the story than the funding announcement.

Rebesco has described the early company as focused on revenue and on customers willing to give honest feedback. A paying customer has a useful freedom: the ability to say the product is not worth its price. That is uncomfortable information for a founder, and potentially valuable information for a product team.

He has said the company first secured an Army contract in 2019. Those early relationships gave development a concrete audience. An organization buying software brings its own workflow, expectations, and objections. The objections can be more instructive than enthusiasm, especially when the goal is to keep a system useful through repeated use.

By March 2026, Striveworks had announced Series B investment led by Washington Harbour Partners. The stated purpose included expanding engineering and research capacity and meeting demand across U.S. and allied governments. For Rebesco, growth meant extending a capability the company had already spent years developing and deploying.

Capital gives a company room to do more work. It also enlarges the practical questions. More deployments mean more environments, more users, and more occasions for yesterday’s assumptions to encounter tomorrow’s complications.

The sea refuses to hold still

The maritime setting makes Rebesco’s preoccupation unusually tangible. Visibility changes. Sea conditions change. A sensor on one vessel does not automatically supply the same view as a sensor on another. A useful model has to meet those differences somewhere beyond a presentation slide.

In April 2026, Striveworks announced a demonstration for the U.S. Navy in which AI operated across multiple vessels as conditions changed. The company described incorporating new operational data, updating models, and transferring them between vessels and sensors across different sea states. Retrained models were delivered on a timetable measured in hours.

The account also emphasized data lineage, the record connecting a model and its results to the information behind them. Rebesco’s concern with continuity appears here in a concrete form: updates have to reach an operating system while people remain able to examine its history.

In a subsequent maritime interview, he discussed the demonstrated work in terms of resilience under environmental and adversary changes. The distinction is useful. A model that recognizes something under one set of conditions has answered an initial question. Keeping it useful when the conditions change creates a continuing obligation.

The person who still approves

Speed creates another question: who has authority over the result? In October 2026, Striveworks announced work in Europe supporting Task Force RAVEN, an Army effort applying AI agents to operational tasks. The described missions included intelligence assessments, target validation, and fire-support coordination.

The company’s account put human review and authorization inside those workflows. Commanders set and update limits. Operators retain approval over major decisions. Chariot Forge coordinates agents across tools and data systems, while logging actions so people can trace what the system has done.

For Rebesco, the announcement was another occasion to emphasize testing, evaluation, and guardrails. An agent moving information between systems creates a chain of actions. Accountability requires people to inspect that chain and retain the ability to control it.

Trust also appears in his account of working with colleagues. He urges people to begin with a generous interpretation of another person’s intentions, then address the disagreement directly. The company’s name draws on Theodore Roosevelt’s arena speech and its emphasis on doing the work. Rebesco joked that his own alternative would have been “Thing Doers Inc.” It would, at least, have made the job description admirably brief.

Tomorrow has another appointment

On October 8, 2026, Striveworks announced a five-year, $200 million Army contract supporting Next Generation Command and Control in the Indo-Pacific. The company said the award would expand common AI tools and services, including the ability to deploy, monitor, and re-optimize models and agents as battlefield conditions change.

That announcement followed April’s $70 million enterprise agreement, described as providing technology access for up to 950,000 eligible defense personnel over two years. Those are announced contract and access figures. They describe the scale of the opportunity and the work authorized, rather than money already earned or a count of active users.

The expanding footprint gives Rebesco more places to test the idea that has traveled through his career. An algorithm needs an operating life: information coming in, performance being examined, mistakes being corrected, and people remaining responsible for its use.

His interviews return to those tasks with the persistence of someone who has found a problem worth keeping. The scientific education, the trading career, and the defense company become easier to connect through that interest. Each setting offers a new version of the same awkward appointment.

Tomorrow will arrive with different conditions. Rebesco is building a business around being ready to work on them.