From Theorist to Tech CEO

It began with a cold email. Traversal, a startup that recently emerged from stealth, got its start when a Columbia Engineering MS student Ahmed Lone reached out to Anish Agarwal, an assistant professor in Columbia Engineering’s Department of Industrial Engineering and Operations Research, asking him about his work.

“He wanted to talk about research, and we just really hit it off talking about life more generally,” Agarwal recalled.

Eighteen months later, Traversal has backing from two of Silicon Valley’s most storied venture capital firms: Sequoia Capital and Kleiner Perkins. For co-founder and CEO Agarwal, the launch is the culmination of a sprint that transformed a theoretical idea into a company now poised to reshape enterprise software.

Using AI to unravel cause and effect

Traversal’s core mission is to keep large software systems running smoothly. 

“Modern software systems are massive, interconnected networks,” Agarwal said. “When something goes wrong, it spreads like an epidemic. You see dozens of red alerts, and figuring out which one is the root cause is incredibly hard.”

Industry stalwarts like Datadog, Splunk, Dynatrace, and Grafana, broadly called observability platforms, store and visualize telemetry data, providing nice dashboards so engineers have a window into the health of their system. But the painful toil of “dumpster diving” through these dashboards when an incident occurs still remains the painful job of on-call engineers. These software failures are becoming ever more frequent with the advent of “vibe coding,” the practice of writing software where the majority of it is generated by AI — in such situations engineers have even less context about their system, making finding the root cause far more challenging.

Agarwal is trying to usher in what he calls observability 2.0, where AI agents obviate the need for on-call for good. Agarwal saw an opportunity to apply his research in causal machine learning — a field focused on uncovering cause-and-effect relationships hidden in large datasets — to this persistent problem. With the rapid evolution of large language models and AI agents, he realized these tools combined with his research in causal machine learning could finally help automate the process of diagnosing and resolving production incidents, something that has eluded academics and practitioners alike for decades.

A Columbia story

The founding team reflects that ambition. In addition to Agarwal, Traversal’s co-founders include Lone, a Columbia master’s graduate; Raaz Dwivedi, a faculty member from Cornell Tech; and Raj Agrawal, who Agarwal met during his PhD at MIT. The company now counts at least eight Columbia alumni among its 35 employees.

Image

“This company has Columbia DNA,” Agarwal said. “I’m grateful for the flexibility and support the university has given me on this journey.”

The startup recently announced $48 million in seed and Series A funding, with Sequoia and Kleiner Perkins leading them. Agarwal credits his success in securing funding for Traversal — his first startup — to its unique position in the market and an evolving technological ecosystem: demand for this kind of product is high, but executing it using newly emerging AI techniques is ever-evolving and requires constantly being right at the cutting edge of what the models can do.

“This is a very hard AI and statistics problem, arguably the most difficult research problem I have ever taken on in my career,” Agarwal said. “But if you can solve it, there is a huge market for it just waiting to be taken.”

Lead Photo Caption: The team at Traversal, from left to right: Raaz Dwivedi, Anish Agarwal, Ahmed Lone, and Raj Agrawal

Lead Photo Credit: Traversal