Loki Cheema Learned Freight From the Hold Music

The first useful piece of freight software Loki Cheema ever supplied was himself. He was sixteen, working at a motel in Everett, Washington, when a long-haul driver needed help with paperwork. They spoke the same language, Punjabi, and Cheema understood enough to make the forms less forbidding. He helped. Then he helped the driver's friends. A small favor became a weekly job, and the lobby became an accidental classroom in American logistics.

The curriculum was not abstract. Bills of lading had to be correct. Logbooks had to make sense. Settlement statements determined whether a driver got paid. A missing detail did not produce an elegant error message. It produced a delayed check, another call and another hour surrendered to a system whose parts could not agree on what had happened.

Cheema was dealing with instability of his own. He has said he was kicked out of home and faced homelessness at sixteen, working punishing hours at the motel. Trucking offered a foothold. He saved enough to move from helping with forms into dispatching trucks and, eventually, fleet operations. The work put a roof over his head. It also gave him an intimate index of the industry's small absurdities.

The truck was never the problem. The information was the problem.Loki Cheema

An education in waiting

At one point, Cheema says, he made more than a hundred calls a day. He could name a trucking company and recite its hold music. It is a funny boast until you consider what it measures: not musical memory, but hours of human attention converted into queue time. He remembers having three phones on hold at once. He saw offices run on paper and Excel, with documents arriving in WhatsApp chats connected to printers. Freight was moving across the country while its information stood beside the copier.

Then came a roadside lesson. Cheema once broke down in a truck in the middle of Arizona. The machinery, the phone queues and the bureaucracy converged into one stranded afternoon. He later wrote that he would gladly spend his life ensuring other operators did not have to endure the same inefficiency. The sentence is dramatic. The underlying product note is wonderfully plain: when software fails in logistics, someone is often waiting in a physical place.

His route did not proceed neatly from truck cab to code editor. Cheema started an ethnic grocery store and restaurant in California. Those businesses supplied another view of logistics, now from the customer end: products to source, inventory to manage, suppliers to trust and people to meet. A restaurant may seem far from a transportation management system, but both punish vague information and late arrivals with unusual speed.

The old jobs become the feature list

By the end of 2024, voice AI had advanced far enough for Cheema to revisit the chores he knew by muscle memory. He sold the grocery store and restaurant to bootstrap Hemut, then began assembling a team. In one post, he compared the process to collecting Infinity Stones, a comic-book flourish attached to the distinctly adult task of recruiting people who can survive a complicated market.

Hemut's premise is that trucking does not need another narrow screen. A carrier should be able to find and book a load, communicate with a driver, process a document, send an invoice and reconcile the books from one operating picture. AI voice agents can handle inbound and outbound calls. Document parsing can remove rekeying. Automations can carry routine financial work forward when a delivery is confirmed.

The distinction is between software that watches and software that works. Traditional dashboards make operations visible, which is useful. Cheema wants the system to take the next permitted step as well. A dispatcher who no longer copies the same detail between tabs can spend that attention on a driver or customer. The technology recedes; the relationship gets the recovered minutes.

This depends on time. A nightly batch can produce a handsome report about yesterday, but trucking margin is decided while the load is moving. “Every hour between the data and the decision costs a carrier real money,” Cheema has said. Hemut's current architecture emphasizes real-time streams so a truck's location, a fault code, a delivery confirmation and the ledger belong to the same present tense.

A signal, then a larger fleet

The team initially rolled out what Cheema called a “mini TMS” for smaller carriers. He reported that it reached $30,000 in annual recurring revenue in two weeks. That was less a victory lap than a signal: operators would pay to retire part of the patchwork. Hemut then moved upstream and signed Meiborg, a fleet Cheema described as nearly 200 trucks. In his June 2025 launch post, he reported $270,000 in annual recurring revenue.

Those are founder-reported early figures, useful chiefly because they reveal the sequence. Start with a narrow version. Watch whether the pain is sharp enough to produce a quick purchase. Then attempt the harder migration, where a larger fleet has more history, more integrations and more people with reasonable objections to changing the machinery beneath their work.

Y Combinator accepted Hemut into its Spring 2025 batch. The credential brought a familiar startup setting to a founder whose useful knowledge came from less photogenic rooms. On the public company page, his biography begins with homelessness, motel shifts and a driver who needed help. The product description nearby lists AI calls, document uploads, load sourcing and accounting. Biography and roadmap sit a few lines apart.

Cheema has also studied computer science and linguistics at UCLA, an interesting pair for a founder working on machines that must interpret people under operational pressure. The disciplines meet in a voice agent answering a status call: software needs the correct data, but it also needs to understand an impatient human asking an imperfect question. His education did not replace his operating experience. It supplied another vocabulary for problems he had already met at the dispatch desk.

The team around him reflects the breadth of the problem. Hemut's public roster spans product, operations, engineering, AI, design, sales and marketing. That range matters because replacing a carrier's patchwork is not merely a coding assignment. Someone has to understand how the existing work gets done, how a dispatcher reacts at 2 a.m., how data travels, how a migration avoids interrupting live freight and why an operator should trust the new system on Monday morning.

Cheema's stated instinct is social as well as technical. He opened a grocery store and restaurant partly to connect with people. Hemut's team page describes logistics as a relationship business, and its product argument repeatedly returns to giving people more time for those relationships. It is a useful correction to the chilly language of “autonomous operations.” In this version, autonomy handles the repetition so a human can handle the exception, the negotiation or the driver who needs an answer with judgment attached.

You get a certain number of years to build something that counts, and I chose this.Loki Cheema

What the software owes the road

Cheema's aspiration has widened since the mini TMS. Hemut now talks about an internet of freight built on real-time data, with carriers connecting to brokers and shippers without manually re-entering the same load. The phrase is large. His explanation remains grounded in the motel lesson: a carrier should run from one live picture rather than five systems learning the truth a day late.

There is restraint hidden inside that ambition. The most important promise is not that an agent can talk. It is that the agent has the right context when the phone rings. It knows the truck's current location and the fault code that just appeared. It can answer without sending another person into another queue. Intelligence without timely operational data would merely provide more eloquent hold music.

Cheema's story offers a practical lesson for anyone building vertical software. Look for the work people repeat while nobody is applauding. Learn why the spreadsheet survives. Notice which phone call interrupts dinner and which missing field delays payment. Product insight often arrives dressed as tedium. The founder's advantage is having been bored, stranded or overworked in precisely the right place, then remembering the details.

Trucking gave Cheema shelter and a trade when he needed both. He speaks about Hemut as a way to return something to its operators and drivers. That debt gives the company its most interesting test. If the software works, the evidence will not be another screen glowing in a back office. It will be the call nobody had to make, the form nobody had to copy, and an evening that once belonged to hold music quietly returned to its owner.