Aman Singla has spent much of his career in the part of technology that disappears when it works. A phone finds Wi-Fi. A home network adjusts itself. A business user asks a question and gets an answer assembled from systems that were never designed to cooperate. The experience feels simple only because somebody absorbed the complexity underneath it.
Singla's current company, MarcoPolo, sits in precisely that underneath. It gives AI assistants a governed workspace for reaching databases, cloud storage, warehouses and business applications. The pitch is less about making a model clever than making it useful inside a real organization, where credentials have scopes, data has owners, actions need logs and the meaning of a field may live in one employee's memory.
This looks like a timely jump into agentic AI. Seen across three decades, it looks more like the latest version of an old habit. Singla keeps finding a promising technology at the moment it collides with operational reality. Then he builds the layer in between.
01 / The systems instinct
Before the product, there was the structure
At IIT Delhi, where he studied computer science and engineering from 1988 to 1992, Singla wrote his undergraduate thesis on graph theory and algorithms. At Georgia Tech, his doctoral work placed him among researchers thinking about parallel processing, computer architecture, distributed systems and approximation algorithms. Those subjects ask a shared question: how do independent parts coordinate under constraints?
His first documented industry stop was Silicon Graphics, where he worked as a senior member of technical staff. In 2000 he joined Atheros Communications, then roughly a 30-person startup making wireless semiconductors. Wi-Fi was moving into phones and consumer devices, which meant software had to make new radio hardware reliable across an unruly world of products and networks.
Singla moved from software engineering director to vice president of engineering for the consumer business. Atheros grew past $1 billion in revenue and Qualcomm acquired it in 2011 for about $3.6 billion. He continued into Qualcomm, eventually serving as vice president of software. The arc gave him an education that no architecture diagram can provide: a system changes when the company around it moves from dozens of people to global scale.
I've always been motivated by building teams around new ideas.Aman Singla, on joining Mayfield as an entrepreneur-in-residence
The quote is revealing because it puts teams before products. Infrastructure work is coordination work. Hardware, software, customers and standards all arrive with different clocks. Someone has to design the interfaces, decide what can fail gracefully and build a group capable of operating the answer.
02 / Moving up the stack
The home router became a cloud problem
In 2015, Singla joined Plume's founding team as co-founder and CTO. The company looked at home Wi-Fi and treated it as more than a box in the corner. The network could become a cloud-managed service, learning from conditions across access points and devices, then adjusting performance through software.
That shift sounds natural now. Operationally, it meant turning a hardware ecosystem into a living data platform. Plume had to ingest enormous volumes of network events, serve internet providers of very different sizes and handle privacy requirements across markets. A published infrastructure case study described clusters processing 27 billion operations a day, with the ability to support 75 billion. Singla is named on patent filings for cloud-based control and optimization of Wi-Fi networks, evidence of the technical work behind the platform story.
When the router moves to the cloud, scale stops being an abstraction.
Operations per day described in Plume's published data-infrastructure case study.
The lesson was not merely that cloud software could improve Wi-Fi. It was that the useful product lived across layers. Radios generated signals. Distributed databases held state. Machine learning found patterns. Operators needed compliance and predictable performance. Consumers wanted the internet to work without learning any of this.
03 / A company hidden inside a problem
Immersa found the deeper bottleneck
After Plume, Singla spent several months at Mayfield as an entrepreneur-in-residence alongside Aseem Chandra. Their stated ambition was to make AI and machine learning useful to business employees who could not summon a Fortune 500 data team. In 2021, the work became Immersa.
Immersa began with revenue operations. Sales and service teams have customer information scattered across product logs, a warehouse, a CRM, support tickets and spreadsheets. The important question is often easy to say and difficult to compute: Which accounts are at risk? Where is usage rising? What should a manager do next? Immersa built data intelligence and automation around those workflows.
The company also encountered a broader problem. Before an AI system can answer a useful business question, it must locate the right data, understand its shape, respect permissions and retain enough context to reason across sources. Every customer-specific integration threatened to become another brittle bridge. The plumbing was beginning to look more general than the application.
The founder's reusable move
When the same hard problem appears behind every customer request, consider whether the obstruction is the larger product.
MarcoPolo emerged from that observation in 2024. The new focus was an MCP-connected workspace where AI tools could query and act on enterprise systems through one governed layer. The team's old problem did not disappear. It became the product.
04 / Context is company memory
Giving an agent somewhere safe to work
MarcoPolo's product thesis begins with a mundane truth: company data is not one neat corpus. It is a Salesforce object, a Snowflake table, a Jira ticket, an S3 log, a permission inherited from an identity provider and a rule that everyone follows but nobody wrote down. A model can be powerful and still fail because it does not know which revenue field finance considers canonical.
The workspace is meant to hold that accumulated context separately from any single AI model. Connections are made once. Credentials remain scoped to a user. Work runs in isolated environments. Tool calls can be audited. The same context can travel across Claude, ChatGPT, Cursor, Copilot or another surface. In this framing, the model is a replaceable component. The company's understanding of itself is the durable asset.
This is also why Singla's older work matters. Agent infrastructure is distributed-systems engineering with unusually capable, unpredictable participants. It needs boundaries, observability and fallbacks. The interface may be a conversation, but the production burden looks familiar to anyone who has watched a network, a cloud service or a database meet real traffic.
At Zuora's AI Week in 2025, Singla spoke about context management as a requirement for effective LLM reasoning, not merely retrieval. By 2026, MarcoPolo was presenting a workspace built around that position and publishing field notes from conversations with enterprise AI teams. The recurring complaints were data access, security and context, not a shortage of impressive models.
The fashionable layer gets the demo. The enabling layer determines whether the demo becomes a system.The pattern across Singla's career
05 / Building institutions, too
A systems view beyond software
Singla's interest in enabling layers extends outside companies. He is part of the founding community of Plaksha University, a technology university in Punjab assembled by entrepreneurs, academics and business leaders. The role fits the rest of his biography. A technical workforce does not materialize when a startup posts jobs. It is produced over years by institutions that connect research, teaching and practice.
There is no need to turn that into a grand theory of one person. It is enough to notice the preference: build the environment in which other people can do difficult work. At Atheros, that meant teams and software around new wireless hardware. At Plume, it meant a cloud platform around home networks. At MarcoPolo, it means a controlled workspace around AI. At Plaksha, it means helping establish a university around technology education.
The aspiration at MarcoPolo is concrete. Let AI work with the systems where business actually happens, without forcing a company to surrender control of its credentials, context or data. Whether that layer becomes a standard part of the enterprise stack will depend on execution, adoption and how quickly the underlying protocols evolve. Singla's career offers no guarantee. It does offer relevant preparation.
He has seen hardware become software, software become a service and a focused application reveal a platform beneath it. Each transition rewarded the same instinct: pay attention to the awkward seam between what technology promises and what operations permit. That seam is easy to ignore. It is also where Aman Singla keeps finding his next piece of work.