Ishan
Guru

An AI That Reads Freight. So Humans Don't Have To.

On a Saturday evening somewhere in America, a trucking company owner used to spend six hours manually entering paper work orders into a computer instead of being home with his family. Over 100 orders. By hand. Every week. That was the specific detail that crystallized what Pallet was for. And Ishan Guru helped build the thing that ended that ritual.

Ishan Guru is Co-Founder and Founding Engineer, Applied AI at Pallet - a San Francisco-based startup whose product, CoPallet, functions as an AI workforce for logistics providers. The platform automates the back-office operations that have consumed the freight industry for decades: order entry, rate quoting, carrier coordination, portal updates, document processing. It does this 10 times faster than a human operator, at less than half the cost.

In May 2025, Pallet closed a $27 million Series B led by General Catalyst - seven months after its Series A from Bain Capital Ventures. Total funding: $48 million. The investors are not betting on a concept. They are betting on math that is already working.

Ishan arrived at Pallet not as a passenger but as an architect. His role as Founding Engineer means he was there when the stack was a whiteboard. With two computer science degrees from Columbia University - one in Software Systems, one in Machine Learning - he is exactly the kind of builder who belongs at the intersection of complex distributed systems and applied AI. The logistics industry happens to sit at that intersection too. It just didn't know it yet.

A Trillion Dollars of Paperwork, Waiting for Software

Global logistics spends more than $1 trillion annually on administrative back-office work. That is three times the size of the entire SaaS market. This is not a niche. It is the most under-digitized sector in the global economy - an industry where fax machines are still operational, where order confirmations arrive as PDFs attached to emails, where a mid-sized carrier might have 25 full-time employees doing nothing but data entry.

The freight industry has legacy software. It has transportation management systems (TMS). What it does not have - or did not have before companies like Pallet arrived - is software that can actually understand the messy, multi-format, exception-riddled nature of real logistics workflows.

CoPallet is not a chatbot bolted onto a spreadsheet. It is an event-driven AI agent system - built by Pallet's engineering team to handle the long-running, distributed, exception-filled reality of logistics workflows. The architecture uses Zod schemas for type safety, OpenTelemetry for observability, and dynamically allocates compute based on task complexity. The engineering decisions are deliberate because the stakes are real: one wrong shipment update can cascade into a missed delivery, a failed audit, a lost client.

From Papua New Guinea to Pallet: An Unlikely Straight Line

Ishan Guru's career does not follow the standard trajectory. Before he was writing machine learning systems at Columbia or shipping payments infrastructure at AWS, he was an intern at Datec PNG Limited in Papua New Guinea - one of the Pacific's leading IT companies - and at City Pharmacy Limited Group, one of the country's largest retail chains. It is an unusual origin story for a New York-based AI engineer, and it matters: he has seen how technology works - or doesn't work - outside the comfortable bubbles of Silicon Valley and Wall Street.

From Papua New Guinea, the path ran through Goldman Sachs, where he joined as a software engineer and summer analyst in 2016. Goldman is not known for producing founders; it is known for producing engineers who understand how complex systems fail under pressure, how data flows through large organizations, and what it costs when software is wrong. These are useful lessons for anyone building AI that touches real-world logistics.

Then came Amazon Web Services, where Ishan served as a Technology Lead Software Engineer. AWS is where you learn to think at infrastructure scale - how distributed systems behave, what reliability actually means when you cannot afford downtime, how cloud architecture shapes everything built on top of it. This background is not incidental to what he builds at Pallet. It is the substrate.

Between AWS and Pallet came Violet - a startup building a universal commerce API that let developers integrate checkout and payments across any platform with a single interface. Ishan rose from Lead, Payments Engineer to Director of Product Engineering, overseeing the architecture of cross-platform commerce infrastructure. The patterns are consistent across his career: complex networked systems, data flowing through multiple parties and formats, the challenge of making disparate things talk to each other reliably at scale. Logistics is just the next domain where those patterns apply.

Columbia, Twice: Software Systems Then Machine Learning

Ishan holds both a Bachelor of Science in Computer Science (Software Systems) and a Master of Science in Computer Science (Machine Learning) from Columbia University. The pairing is deliberate, not accidental. Software systems thinking - how distributed applications are architected, how data flows, how reliability is engineered - combined with machine learning - how models are trained, evaluated, and deployed in production - is exactly the skill set required to build AI that actually works inside messy real-world operational environments.

Logistics is precisely that: a messy real-world operational environment. Orders arrive as PDFs, Excel files, handwritten notes, EDI messages, and email text. Each carrier has its own portal format. Each shipper has its own process. The AI that Ishan helps build at Pallet must handle all of it, learn from corrections, and improve - without a PhD in freight required from the human on the other end.

CoPallet: The AI That Goes to Work So Freight People Don't Have To Work Twice

CoPallet is Pallet's flagship AI platform. It automates the manual back-office operations that consume logistics companies - order entry from multi-format documents, rate quoting, load tendering, carrier coordination, portal updates, real-time shipment visibility, document digitization, and more. One mid-sized carrier that adopted CoPallet was able to reallocate 25 employees who had been doing repetitive order entry - saving the company millions of dollars annually.

The product handles edge cases. That is the crucial detail. Any logistics workflow automation built on rigid rules breaks the moment an unusual order format appears, or a shipment spans multiple legs with different carriers, or a document arrives in a format nobody anticipated. CoPallet uses a human-in-the-loop model - AI handles the standard volume automatically; humans review and correct exceptions; the system learns from every correction and gets better at handling that exception next time. This is what Pallet means by an "AI feedback loop" and "workflow learning."

The engineering team at Pallet built CoPallet on an event-driven architecture specifically because logistics workflows are long-running and distributed. An order might take hours or days to fully process across multiple parties, systems, and checkpoints. That is fundamentally different from a typical web application that responds in milliseconds. The architecture reflects deep understanding of the problem domain - the kind of understanding that comes from engineers who have spent time at companies like AWS and Violet, building systems designed to handle real-world complexity at scale.

Pallet's AI agents have four core capabilities: procedural knowledge (workflows), context retention (memory), system interfaces (tools, via APIs, browser automation, and computer vision), and logical problem-solving (reasoning). They are not scripts pretending to be AI. They are AI systems built to handle the infinite variability of physical goods moving through the real world.

$48M Raised, 130 People Strong, and Just Getting Started

Pallet was co-founded by Sushanth Raman (CEO) and Andrew Spencer, both former engineers at Retool, with Ishan Guru joining as Founding Engineer and Co-Founder to lead the applied AI engineering work. The company is headquartered in San Francisco, with Ishan based in New York - a geographic footprint that reflects where the logistics industry lives: not just in Silicon Valley, but in the ports, depots, and freight corridors of industrial America.

The investor roster reads like a who's-who of the best-performing enterprise software funds: General Catalyst led the Series B, with participation from Bain Capital Ventures, Activant Capital, Bessemer Venture Partners, Vicus Ventures, and BoxGroup. Marc Bhargava of General Catalyst called it "a multi-billion dollar opportunity." At $1 trillion in global logistics back-office spend as the addressable market, that assessment is conservative rather than generous.

The raise comes as tariff uncertainty and global supply chain pressure force logistics operators to cut costs fast. CoPallet is the answer that requires no headcount reduction announcements - just reallocation, from manual repetition to actual judgment work. For a 130-person company that has raised $48 million and already has enterprise customers showing 90% employee adoption rates and 70% workflow reductions, the trajectory is clear.

Compassionate Engineer, Deep-Water Diver, Global Citizen

Colleagues describe Ishan Guru with words that are rare in engineering profiles: "compassion," "positive energy," "makes everyone else around him better." These are not the descriptors typically attached to machine learning engineers building distributed systems. They are the descriptors of a leader who understands that software is made by people, for people, and that the culture of the team shapes what the product becomes.

Outside of freight AI, Ishan holds a PADI Advanced Open Water Scuba Diver certification. There is something fitting about that: the willingness to descend into environments that are genuinely complex, where visibility is limited, where you have to trust your equipment and your training, and where the rewards come from understanding systems that most people never see. Building applied AI for logistics is not so different.

The Papua New Guinea chapter of his story is the one that most people skip past but shouldn't. Working in the IT and retail sectors of a Pacific island nation, far outside the cloud-first assumptions of American tech, teaches things that no amount of AWS certification can. It teaches you that the world is not flat, that infrastructure is not reliable everywhere, that the people who actually need technology to work are often the ones building it by hand. That perspective shapes how you build AI for industries that have been stitched together with manual labor for decades.

The New Operating System for Physical Goods

Pallet describes its mission as building "a modern OS for moving any physical product from point A to B." That is not a metaphor. It is an engineering target. The endgame is logistics that are as frictionless as modern software deployment - where an order enters the system, is understood by AI, coordinated across carriers and shippers, tracked in real time, and delivered with no human having to manually touch a keyboard to make any of it happen.

The $27 million Series B will fund infrastructure scaling and continued CoPallet development. The logistics automation market is not seasonal: tariff disruption, e-commerce growth, near-shoring trends, and supply chain resilience investment all push demand in the same direction. Pallet is positioned not as a single-feature tool but as the back-office layer that every logistics company eventually needs.

Ishan Guru's role in that future is the one he has been training for since he first learned to think in distributed systems at Columbia, then at Goldman, then at AWS, then at Violet: building the infrastructure that makes complex coordination feel simple on the surface. The freight industry has waited a long time for software this good. The wait appears to be over.

Go Deeper