The trouble began with a hospital bill. After an emergency department visit, Preeti Bhargava received a charge that seemed wrong. Months of exchanges with an insurer and provider eventually exposed a coding error. The medicine had happened; the paperwork had misrepresented it. In Arintra’s account of its origins, that small administrative failure helped inspire a company. It also supplies a useful entrance into the thinking of one of its investors, FoundersX Ventures.
- Invest across AI’s machinery, models and industry applications.
- Combine technical research with early venture investment.
- Look for a specific problem before selling a sprawling platform.
A hospital coding desk seems an unlikely neighbor for a quantum computer. Yet both appear in the FoundersX portfolio. So do SpaceX, World Labs, cybersecurity company Salt Security and patent software business Solve Intelligence. The interesting question is what connects them. For FoundersX, the answer is a view of AI as a chain of dependencies: better infrastructure makes new capabilities possible; specialized businesses turn those capabilities into work somebody needs done.
A rocket and a reimbursement claim
Founded by Helen Liang in 2016, FoundersX invests across what it calls the full AI stack. That includes chips, space infrastructure, physical AI models, robotics and agents operating inside industries. Think of the portfolio as a cross-section through a machine. Some companies supply its underlying capacity. Others teach systems about the physical world. At the customer end are tools that address billing, security, finance or intellectual property.
This framing distinguishes its pitch from a fund devoted only to enterprise software or only to hardware. It also creates a demanding research assignment. An investor evaluating a quantum architecture cannot simply reuse the checklist for a hospital software purchase. The proposed advantage is understanding how changes in one layer create opportunities in another. That is an investment thesis, with all the uncertainty the phrase implies.
The reading room behind the checkbook
Liang’s public biography combines a Wisconsin-Madison PhD with more than a decade of high-tech product and operations experience. Partner Tom Kosnik brings a history of teaching technology ventures at Stanford and Harvard Business School. General partner Leo Cui has a Harvard bioengineering PhD and postdoctoral experience in Feng Zhang’s lab. General partner Sizi Chen studied music before earning a Stanford master’s in Management Science and Engineering and working at the MIT Media Lab.
Those biographies make the breadth less mysterious. The team has people accustomed to moving between disciplines. Liang says its routine includes a deep research topic every other week, ranging from multi-agent systems to quantum chips and energy infrastructure. Her compact explanation of the approach is memorable:
“we don’t win by capital, we win by brains”Helen Liang · 2026 AGM recap
The firm also advertises a year-long fellowship with actual investment work: analyses, diligence, sourcing, portfolio calls and AI research. For students, it offers an entry into venture practice. For the firm, the structure puts university and research networks close to the investment process. Repeated study becomes part of the operation rather than an occasional conference excursion.
Earn the next workflow
Arintra makes the commercial logic concrete. Its software interprets clinical documentation and automates medical coding within existing health-record systems. In its August 2025 financing announcement, Arintra reported that Mercyhealth saw coding costs fall 32% and denials on automated claims fall 43%. These are company-reported customer results, with a specific setting and denominator. They illustrate what a buyer can evaluate: fewer rejected claims, less manual work and faster payment.
Arintra-reported results at Mercyhealth, August 2025. Outcomes are specific to this deployment.
In a November 2025 healthcare essay, FoundersX describes the sequence explicitly: begin with a reliable, narrow task, deliver economic value, then expand into a broader platform. Solve Intelligence offers another version. FoundersX says it backed the company at seed. Solve began with patent drafting and prosecution; its December 2025 announcement paired a $40 million Series B with Charts, extending into claim charts and patent analysis. The next product follows an existing professional workflow.
Two clocks, one portfolio
There is capital behind this patience. FoundersX’s current website reports about $800 million under management, 12 unicorn investments and 10 acquisitions. Those figures describe the firm’s stated scale and portfolio milestones. They do not tell a reader how much cash investors have received. A billion-dollar valuation is a price attached to a company at a moment in time; an acquisition count says little about proceeds without ownership and transaction details.
The business itself is private-market investing: manage investment vehicles, take stakes and support companies as they grow. FoundersX says it leads seed deals and co-leads Series A rounds. Its audiences are fund investors seeking venture exposure and founders seeking capital, talent connections and introductions. A hospital buying Arintra is a customer of Arintra. Keeping those relationships straight matters when judging what the venture firm actually sells.
The two ends of the portfolio also keep different time. FoundersX says it first backed Universal Quantum in 2020; the company announced a Series A exceeding $100 million in October 2026. Software adoption and physical engineering cannot share an automatic timetable. The portfolio asks investors to accommodate both. Its 2025 Manhattan office expansion, alongside Silicon Valley and Cambridge, places the team nearer another set of founders and research communities.
The useful part to steal
A founder can borrow the method without borrowing the portfolio. Pick a painful task. Identify who owns its budget. Measure the current delay, error rate or labor bill. Show that a proposed tool fits the customer’s systems, then establish what improvement justifies payment. Expansion becomes more credible when the first job works. A grand market forecast cannot do that work for you.
The conditions matter. If data is inaccessible, outputs are unreliable or integration consumes the savings, the narrow entry point may stay narrow. Hardware adds manufacturing and financing constraints that a software sales cycle cannot resolve. FoundersX’s broad thesis is most useful when it sends a builder back to these particulars. Somewhere between the chip and the reimbursement claim sits a person deciding whether the machine is worth buying.