A tennis scoreboard is a brutally economical editor. It does not care how convincing the warm-up looked, how elegant the plan sounded, or how good the previous point felt. It keeps two numbers and advances them without sentiment. Marshall Leung spent four college seasons learning inside that arithmetic. Now he works in a world that has the opposite problem: industrial operations produce oceans of numbers, scattered across warehouse, transport, enterprise and labor systems, while the person running the floor still needs one useful answer.
Leung is part of the founding go-to-market effort for Neo, Traba's decision-intelligence product for supply-chain operators. The role puts him between what software can do and what a warehouse or manufacturing operator needs before the next shift gets expensive. It is early-career work conducted in grown-up conditions: late shipments, missed billing, overtime, inconsistent system records and the daily opera of five browser tabs disagreeing before breakfast.
The path there is short enough to see clearly. Leung came to Amherst College from Tiburon, California, and Laurel Springs School. He joined the men's tennis roster, eventually captained the team, won a regional B singles title and earned an all-academic conference honor. Then came New York and Traba. The court changed. The underlying assignment did not: read what is happening, decide what matters, and make the next move before the opening closes.
A championship built from untidy points
The neat version of Leung's best documented college result is one line: B singles champion at the 2022 ITA Division III New England Regional Championships. The match sequence is more revealing. He opened with an 8-4 win. In the round of 16 he won 6-2, 7-6. The quarterfinal wandered through three different moods, 6-1, 4-6, 10-3. His semifinal required two tiebreaks. Only then came the clean final, 6-3, 6-0.
The temptation is to make sport carry more metaphor than it can bear. Tennis does not automatically manufacture good operators, and a forehand offers limited guidance on enterprise integrations. But the record shows Leung repeatedly working through altered conditions. A set slips away; a match tiebreak follows. Two sets tighten; both go to breakers. The final score becomes simple only after the complicated rounds have been handled.
There were other concise lessons. In March 2024, against Skidmore, Leung won seven consecutive games to finish a 6-4, 6-0 match in under an hour. A month later, in Amherst's 5-4 win over fifth-ranked Bowdoin, his 7-5, 6-4 singles victory supplied one of the team's five points. He also won in conference victories over Hamilton, Colby, Trinity and Bates that spring. The results present momentum as something less mystical than sports language usually makes it. Seven games are still seven separate games.
Momentum is repetition that has finally become visible.On seven consecutive games, and the work between scoreboards
Leung's college record was not confined to winning. He appeared on the 2023 NESCAC Spring All-Academic Team, and his public profile identifies him as an Amherst captain. Neither distinction supplies a personality diagnosis, nor should it. They do establish the two demands he was asked to meet in public: perform in a competitive lineup and maintain the academic standard attached to the honor.
The morning with five answers
Neo begins from a distinctly less elegant scorekeeping system. A third-party logistics operator may have orders in a warehouse management system, costs in an enterprise resource planning platform, labor information in an HR system and shipment data in a transport platform. Each system can be correct on its own terms while the operation remains confused. By the time a supervisor discovers that a line has been running below pace, the overtime bill may already be composing itself.
From scattered systems to an approved action
HRIS · TMS
reconciles + explains
reviews + approves
Traba introduced Neo in June 2026 as an intelligence layer across those systems. It answers plain-language questions with charts, reconciles numbers before presenting them, looks for operational trouble and drafts actions such as claims, order fixes and customer updates. The human remains in charge of what goes out. This matters because industrial AI meets reality at a loading dock, not in a slide deck. A recommendation is useful only when somebody can trust the inputs and act before the cost hardens.
The product grew from Traba's earlier work in light-industrial staffing. The company says the team behind Neo has supported more than one million shifts across more than one thousand facilities. That history gives its software thesis a pair of work boots: the coordination problem was first encountered through people, schedules, attendance and the daily unpredictability of physical operations.
For Leung, go-to-market work means giving that history a useful present tense. A product can connect WMS, ERP, HRIS and TMS data; a buyer still has to understand what changes on Tuesday morning. The bridge is built from specific questions. Can the system show cost per package last week? Can it catch an overcharge? Can it explain why a shipment is at risk? Can it prepare the correction while leaving approval with the operator?
His own public language stays close to that floor. In September, sharing a conversation about Traba, Leung invited supply-chain leaders to a New York breakfast to discuss where AI was “actually making an impact on the floor.” The adverb is doing honest work. AI has had plenty of attention in the abstract. A warehouse has little use for abstraction when headcount and order volume have stopped speaking to each other.
“Traba's growth is a testament to the amazing product and team behind it.”Marshall Leung, September 2026
The useful distance between a pitch and a point
A founding go-to-market role is sometimes dressed up as the art of making noise. The less theatrical version is a cycle of contact and correction. An operator describes a problem. The product either answers it or fails to. The team learns where the explanation was vague, where the workflow pinched, and where the promised value did not survive contact with the shift. Then it returns with a better answer. The work resembles a rally because the ball keeps coming back.
Leung arrives at that job with a public record full of feedback loops. College tennis published his wins, losses, scores, lineup positions and team results. There was no need to speculate about whether the day went well. The numbers were impolite enough to say. Industrial software offers more numbers but often less clarity. Neo's proposition is to restore the useful part of a scoreboard: not simplicity for its own sake, but enough shared truth to decide.
There is a modest comedy in the transition. Tennis gives two people an entire rectangle and asks them to keep one ball in play. Supply-chain software gives hundreds of people several systems, thousands of orders and a margin target, then acts surprised when somebody opens a spreadsheet. Leung has moved from the cleaner problem to the messier one. The advantage is not that he knows every answer. It is that his job begins with discovering which answer the operator can use.
His career remains too young for a grand retrospective, which is a relief. The verified story is more interesting at its present scale. A player from Tiburon joined Amherst, ground through a five-round regional title, met the obligations of an all-academic athlete and captained a team. Two years after his final college season, he was in New York helping introduce a product intended to find costly trouble before it reached the floor.
The continuity is attention under consequence. On court, a late read sends the ball past you. In a warehouse, a late read turns delay into labor cost, missed billing or an awkward call to a customer. One consequence is instantaneous and visible. The other hides in systems until somebody brings it forward.
Leung's work now sits inside that act of bringing forward. He has to help make an unfamiliar product understandable, connect it to a buyer's actual operation and resist the temptation to sell the future when the present contains plenty of unpaid invoices. The score will not be one regional trophy or a handsome launch announcement. It will be whether operators return, whether the software catches what they missed, and whether the next morning requires fewer tabs and less prayer.
A good point ends, and the next one begins almost rudely. There is no ceremony between them. That may be the cleanest way to read Marshall Leung's route so far: not as a leap from athlete to operator, but as a succession of points whose surfaces happen to change. The interesting part is still the decision in front of him.