PineGap.ai wants Wall Street to get its thinking time back

In late 2023, Deepak Sharma asked Ankit Varmani about equity research. Varmani supplied assignments, the sort that keep junior analysts working until midnight. Sharma tried the language models then becoming useful. In Varmani’s account, the discovery was prosaic: much of the work consuming an analyst’s day could be automated. Before there was a company to pitch, there was homework to finish.

That is a more interesting beginning than the usual promise to put artificial intelligence into finance. Which parts require someone to exercise judgment? Which parts require someone to find the same kind of document, extract the same kind of detail and arrange it in the same familiar order? A job can be intellectually demanding while containing a surprising amount of clerical repetition.

Homework becomes a company

PineGap.ai was founded in 2024 by Sharma, now CEO, and Varmani, its Chief Business Officer. Both graduated from IIT-BHU. The company is headquartered in New York, with a technology centre in Bengaluru. Its team brings together former equity researchers and AI engineers. Varmani’s background includes JPMorgan and a long/short equity hedge fund; the product has a practitioner on the premises.

The first disclosed cheque was a $2.5 million seed round in April 2024, led by Silicon Valley Quad and Inventus Capital, with DeVC and angel investors participating. It was intended to accelerate development and engineering hiring in Bengaluru and the US. The more revealing test would come when investment teams had to use the software in their own routines.

The fund supplies the instruction manual

The website presents research as a set of assignments. A Company Primer gets an analyst acquainted with a business. Proxy Review covers board structure and management compensation. Risks Tracker looks for forensic accounting signals. Other workflows address earnings, conferences, deeper company research and idea generation. The vocabulary is refreshingly close to an analyst’s to-do list. Reading a proxy needs no new verb.

The company’s broader offer is customization. It works with fund teams to understand their research processes, investment theses and workflows, then builds agents around their investment style, private data and output formats. A long/short hedge fund, a long-only mutual fund and a registered investment adviser can all need company research without needing the same report. The document’s destination helps determine its design.

This is where a seemingly small distinction matters. PineGap says its agents deliver outputs on a schedule or in response to market events. Research arrives in the analyst’s inbox. A question does not have to be remembered and typed each time. Recurring work can have a recurring instruction, much as a capable colleague can learn what belongs in tomorrow morning’s note.

The investor who expected a DIY answer

Stellaris Venture Partners, which led the Series A, describes entering with doubts. The underlying information was largely public. Funds could plausibly combine foundation models with data connectors and build something themselves. Busy institutional customers would also be difficult to win. These are sensible objections: access to an ingredient is seldom enough to make a durable business.

What changed the investor’s view, according to its published account, was talking to customers. Analysts described a product that fitted their work; some funds with substantial internal AI capabilities still preferred PineGap for these workflows. Stellaris gives a concrete example: a company primer reorganized around one fund’s ten proprietary investment principles. The interesting achievement is that the software can accommodate someone else’s habits.

“Be as frugal as possible until you achieve product-market fit.”Ankit Varmani · August 2026

There is a lesson here that travels beyond financial research. Start with a finished piece of work somebody needs. Identify its inputs, its timing and its intended reader. Only then decide what to automate. A generic summary can be impressive in a demonstration and inconvenient at a desk. The useful test is whether the recipient can pick up the output and continue working.

50,000 reports are a workload, not a return

By August 2026, PineGap reported more than 1,000 deployed agents across over 100 institutional clients, producing upwards of 50,000 research reports each month. Varmani dates the commercial launch to early 2025. The numbers suggest a product being used repeatedly, rather than occasionally admired. They also deserve careful reading: a report count measures activity. It does not tell us whether a portfolio outperformed.

The $8 million Series A brought disclosed funding to $10.5 million. Stellaris led, joined by Inventus, Silicon Valley Quad and DeVC. PineGap said it would expand sales and engineering and build an internal team of former equity research analysts. Hiring researchers to help build research automation is an instructive expense. Understanding the assignment remains part of producing the answer.

The terminal is already learning new tricks

PineGap operates in a market where incumbents have their own AI ambitions. Bloomberg offers conversational research through ASKB, document analysis and AI summaries within its Terminal workflows. The competition is therefore broader than a pile of PDFs and an exhausted analyst. PineGap must earn its place through the fit of its fund-specific work, while customers judge which tools deserve space in an already crowded routine.

Its business is institutional software sold through subscriptions and agreed order forms. Pricing follows the customer agreement. That makes procurement part of the work: a fund has to define what it wants delivered, who will use it and how it will assess the result. More output is only valuable when it improves the next step in the process.

Give the machine an assignment. Keep the judgment.

The arrangement depends on inputs and review. A vague thesis gives a tracker a vague target; an unreliable dataset supplies unreliable raw material. PineGap’s terms say customers retain ownership of their data and that training on it requires prior written consent. They also place responsibility for evaluating outputs and making decisions with the customer. The professional reading the report still has a job.

The practical move is modest: choose one recurring workflow, specify the useful output and compare it with the work your team already trusts. Count the correction time as well as the time saved. PineGap’s founding insight is appealing because it began at that scale. Someone had an assignment. Someone else asked how much of it really needed to keep a human at the desk until midnight.