Alex Li and the Trouble with a Perfectly Good Chart

Smule served 50 million monthly active users. Inside the company, understanding their behavior involved a surprising amount of manual work. Alex Li, its chief technology officer, watched product teams prepare data and pass chart screenshots through long email threads. An app built around musical performances had given its engineers a less tuneful assignment: making sense of the activity behind them.

That irritation became a company. Li founded Kubit in 2018 after years spent building consumer products and the systems underneath them. His recurring concern is the distance between a useful question and an answer a team can inspect. Follow his work from mobile apps to cloud warehouses to AI agents, and the same practical objection keeps surfacing: why should understanding a product require so much clerical work?

The question has acquired new machinery. Kubit now connects what an AI agent does with what a user does afterward. But the founder’s starting point remains recognizably that of an engineer who has sat on the receiving end of an unsatisfactory tool. A beautiful dashboard has limited charm when somebody still has to explain where the number came from.

A career beneath the buttons

Li studied computer science at Shanghai Jiao Tong University and earned a master’s in computer science at Columbia. His career took him through several kinds of software business before he started Kubit. He was a founding member of eBay Classifieds, a principal architect at Jasper Wireless, and vice president of engineering at Booyah. His work crossed marketplaces, connected devices and mobile entertainment.

Those roles gave him a view of products from below the surface. An architect works on the arrangements that allow the visible experience to function. An engineering leader has to keep that experience working as more people arrive. Li eventually spent seven years leading technology at Smule, where the growth of a music app brought a corresponding expansion in data and operational demands.

His public project history includes the launch of Kijiji in September 2005 and a redesign of My eBay earlier that year. These are useful details in a career now described largely through analytics: Li had been building places where people actually clicked, returned, bought and participated. The behavior later represented in reports was attached to products he had helped put into the world.

Colleagues’ recommendations offer a more personal glimpse. Stephen Lim describes an architect who helped other developers work productively and acted as a mentor. Sunit Kavathekar recalls both technical skill and someone enjoyable to work with. Such recollections cannot account for a whole personality. They do, however, place collaboration alongside architecture in the picture of Li that emerges from his working relationships.

The screenshot problem

At Smule, the scale was substantial: Li described handling data from 50 million monthly active users. His account of the difficulty is refreshingly unglamorous. Teams put considerable effort into preparing information, asked questions manually, and communicated findings with pictures of charts. There was plenty to measure, and plenty of labor between measurement and understanding.

A screenshot is a peculiarly final way to share an unfinished investigation. It preserves the picture while leaving the next person to ask about definitions, filters and the reasoning behind it. Li’s early description of Kubit included a workspace and evidence board intended to let colleagues share their thought process. The social life of an analysis mattered alongside the query that produced it.

He also built an internal analytics tool at Smule to replace Mixpanel. That detail gives Kubit’s origin its practical weight. Before there was a business selling an alternative, there had been an engineering team needing one. Li’s frustration concerned both the capabilities of the available software and the work its users inherited.

When Kubit emerged, AI already featured in his account of the product. In 2019 he described automation, anomaly detection and diagnostics as ways to reduce repetitive investigation. His ambition was to let more product people explore engagement and retention without requiring an expert at every turn. The arrival of conversational AI would later change the interface; the wish to remove repetitive work was there early.

Leave the data where it lives

Kubit’s architectural argument developed around the cloud data warehouse, the central repository where a company keeps its business data. Rather than require another separate copy for product analytics, Kubit offered a way to work with the customer’s existing warehouse and data model. Li put ownership and transparency near the center of the proposition.

There is a quiet organizational argument inside that technical choice. When product and data teams work from different stores or definitions, even an ordinary discussion can begin with reconciliation. Li’s answer was to build self-service exploration on the model the customer already controlled. The warehouse supplied a shared foundation; the analytics interface made it accessible to people asking product questions.

In his writing on self-service analytics, he returned to waiting: product teams waiting for analysts, nontechnical colleagues waiting for access to an investigation, decisions waiting for a report. His preferred arrangement gave teams more room to ask follow-up questions themselves. A useful answer should open the next question without sending its author back to the end of a queue.

On April 12, 2022, Kubit announced an $18 million Series A led by Insight Partners, with existing investors Shasta Ventures and TSVC participating. The announcement named Paramount/ViacomCBS, Quizlet and Wish among its enterprise customers. For Li, this moved a complaint formed inside a consumer app into a proposition other companies were prepared to buy and investors were prepared to finance.

“Self-Service analytics must be established upon the data model controlled by the customers.”

Alex Li · April 2022

The funding was a company milestone, and Li’s explanation stayed close to the underlying design. Customers should retain authority over their data model. In June 2023, Kubit announced its platform powered by Snowflake, extending that approach within a major cloud data ecosystem. Li emphasized the bridge between product and data teams, including the ability to continue an investigation with SQL, Jupyter or machine learning.

A founder who hands the credit around

Li’s public explanation of Kubit’s progress includes a revealing distribution of credit. Reflecting on the company’s development, he named engineers, project managers, salespeople, customer success staff and marketers. He attributed its successes to the team. For a founder whose product tries to connect different groups around shared evidence, the list is a fitting one.

He described the mission as helping product and data teams reach better insights faster, tying it directly to his own earlier frustrations. He also welcomed customers including TelevisaUnivision, BET+, Firework and Influence Mobile. The account was grounded in a familiar working problem: a growing enterprise has more information to consider, while the people responsible for decisions still need an intelligible way through it.

His governance writing adds another dimension. He argues for direct warehouse queries, audit trails and data masking so that access and control can coexist. These are the administrative details that rarely make an exciting product demonstration. They matter when an analyst’s useful experiment has to become a process that a larger organization can trust.

These choices suggest a professional interest in making complicated work easier to share. Across his writing, Li returns to the team: who can ask, who can verify, and who must otherwise stop what they are doing to help. The analyst and the product manager need a way to continue the same conversation.

Then the questions learned to talk

On August 7, 2025, Kubit announced Ask Kubit, a conversational interface for customer journey analytics built on its Lumos AI foundation. Users could ask questions in ordinary English. The promised output included an explanation of the query logic, keeping a route back from the answer to the reasoning behind it.

“At Kubit, we believe analytics should be as easy as asking a question.”

Alex Li · August 2025

The sentence is a compact expression of an ambition that reaches back to Smule. Natural language removes one kind of obstacle: knowing how to express the question in a tool’s particular grammar. Explainability addresses another: deciding whether the result deserves confidence. Li’s approach joins the convenience of asking with the practical necessity of checking.

Building AI features brought his team another version of an old annoyance. Li described examining backend agent activity in observability tools and user behavior in product analytics tools, then matching timestamps manually. The instruments recorded different parts of the experience. The person investigating a failure was left to assemble them.

Kubit’s August 2026 Product Hunt launch presented a platform connecting agent actions with user behavior. Documentation updated by Li in September describes a shared event model and user identity across those activities. The aim is to follow an interaction through an agent and onward into the product, so a team can examine the relationship between the AI experience and what happens next.

That brings Li back to the gap that started the business. The objects have changed from mobile events and emailed charts to agent traces and conversations. His team’s complaint still involves a human carrying context between disconnected systems. There is an appealing stubbornness in that continuity. Software has become considerably more talkative; Li is still asking whether it helps its users reach an answer worth acting on.