Most product leaders in artificial intelligence are working at the application layer. They are designing the consumer-facing chat interfaces, the agent workflows, the document generation tools, the enterprise copilots that have absorbed most of the venture capital and most of the attention since 2023. Mr. Palash Kala has spent a decade at the layer beneath them. The decision to keep working at infrastructure rather than building above it has been deliberate.
Suwon, ONOS, and the First Lesson in Carrier-Grade Work
Mr. Kala’s first professional role was at Samsung Electronics in Suwon, South Korea, where he joined the Networks Business as a software engineer in 2016 after a prior internship that earned him Samsung’s best-project award and a pre-placement offer. He had received his Bachelor of Technology in Computer Science from the Indian Institute of Technology, Bombay, one of India’s most selective technical institutions.
At Samsung, he worked on the company’s contributions to the Open Network Operating System (ONOS), an open-source software-defined networking controller hosted by the Open Networking Foundation and the Linux Foundation. Over the period of his tenure, he committed twenty-four changes to the ONOS codebase, the second-highest contribution count among more than thirty developers on Samsung’s ONOS team. He led Samsung’s first Technical Steering Team proposal to the ONOS community, designing and shipping a high-availability capability for network devices managed through the Netconf protocol. The proposal was the first of its kind submitted by Samsung, and its acceptance established the company as a credible voice in shaping the technical direction of the ONOS project rather than only a consumer of the platform.
“Carrier-grade software has a discipline most other software does not,” Mr. Kala has said of that period. “What you ship is going to be running for years on hardware you cannot touch, serving customers who will notice every defect. You learn to think very carefully about every architectural decision, because the cost of being wrong is not abstract.”
He went on to develop network device auto-configuration capabilities, including new ONOS applications and southbound drivers, and a Quality of Service feature for network slicing that was directly aligned with Samsung Networks’ 5G product strategy. The work sat at the foundational layer of the modern mobile internet, where any defect at the platform layer translates directly into degraded service for end users.
Bangalore, worxogo, and Behaviour-Design Infrastructure
In 2019, Mr. Kala moved into product management at worxogo, the Bangalore-based enterprise SaaS company that builds behavioural nudges for sales and frontline organisations. The company had recently closed its Series A from Inventus Capital and Ideaspring Capital and was deploying that capital to expand its enterprise footprint into the United States and Southeast Asia. He joined ten months into the post-funding expansion, and his role quickly extended beyond conventional product management.
Among his most consequential contributions at worxogo was the design and launch of a new product line targeting the senior management tier at enterprise customers. The company’s nudges had primarily served frontline sales representatives and their immediate managers, and the senior leadership tier had been a meaningful but largely unaddressed audience. He took ownership of building the product from scratch, beginning with primary research with leaders across multiple customer organisations, defining the behavioural signals that mattered most for senior decision-makers, and shipping a product that delivered personalised nudges to that layer.
Equally significant was the data and experimentation infrastructure he built. Recognising that the company’s commercial value rested on whether nudges were measurably effective, he conceived and shipped an experimentation tool that the behaviour-labs team used to test which nudge formats, timings, and contexts produced the strongest outcomes. He also designed the data pipelines that powered the company’s analytics dashboards. The internal infrastructure shifted the organisation from reasoning about nudge effectiveness through inference to reasoning about it through controlled experiments.
CRED, and the Discipline of Choosing What Not to Ship
From worxogo, Mr. Kala joined CRED, the Bangalore-based consumer fintech company, where his work materially shaped two of the company’s most visible payment surfaces. He led the launch of CRED Scan and Pay, the company’s scan-based merchant commerce product, which scaled to roughly one million transactions per week and earned a Net Promoter Score of fifty-four at launch. He drove the stabilisation of CRED Pay’s payment success rate from sixty-two per cent to eighty-five per cent at key merchants. The CRED tenure, he has said, was where he learned what a high-trust, high-craft consumer product organisation looks like in operation.
Composio, and the Bet on the Agent Infrastructure Layer
In September 2025, Mr. Kala joined Composio as Head of Product. The agentic AI infrastructure company, founded in 2023 by IIT Bombay alumni Soham Ganatra and Karan Vaidya, is one of the most-watched names in the agent tooling category. Its platform gives AI agents reliable connections to more than a thousand business applications, handling the authentication, rate-limit, and execution-reliability work that, as the company’s founders have argued in public, every individual agent team would otherwise have to rebuild for itself.
The bet underlying the platform is one Mr. Kala has internalised. Most agents demonstrate well and ship poorly, and the gap between the demo and the deployment lives in the unglamorous middle of the stack: authentication, edge cases, schema mismatches, the failure modes that surface at three in the morning when an underlying API changes overnight.
“Infrastructure decisions either enable or constrain the products developers build with you,” Mr. Kala said. “Most of the time they constrain quietly. You have to be very careful about what you ship, because you ship it forever. At the same time, you cannot be precious about it. The agent ecosystem is moving too fast to design every primitive in a vacuum. You have to ship, listen, and iterate in public.”
A Through-Line
Mr. Kala has written about his approach to building products at infrastructure layers. The connective tissue across his decade of work is consistent. At Samsung, the work was carrier-grade software-defined networking. At worxogo, the work was the data infrastructure underneath behavioural nudges. At Composio, the work is the integration and learning layer beneath AI agents.
“The product job is to compress complexity into something a developer can pick up in minutes and trust in production,” Mr. Kala said. “If we get that right, AI agents stop being a 2025 demo and start being a 2026 default. That is the bar we are working toward.”