He read about arthritis biomarkers and photonics back to back, kept seeing the same ideas wearing different lab coats, and built a search engine to chase the overlap.
James Reilly runs Latent Knowledge, a small New York company with an outsized argument: research discovery is broken, and the fix is not a faster index but a smarter map. Its product, LitView, takes the document you are already working on - notes, a half-finished dissertation, a grant draft - and hands back the literature that actually relates to it, arranged as 3D clusters of connected concepts rather than a ranked list of blue links.
The pitch is deceptively simple. Scholars today bounce between separate databases, comb reference lists by hand, and re-run the same queries on five platforms to find a handful of useful papers. Reilly's bet is that a search engine should do the connective tissue for you - and that the most valuable papers are often the ones sitting one discipline over, invisible to a keyword you would never think to type.
Microsoft agreed enough to sign a co-selling deal in 2022. Harvard's Graduate School of Education, the University of Geneva and King's College London agreed enough to put it in front of their researchers. The United States Military Academy at West Point - where Reilly still works as a life-science researcher - agreed enough to roll it out first.
I would be reading about arthritis biomarkers, and then a project on photonics, and it would strike me how many overlaps there were between the two.- James Reilly, on the spark behind LitView
Reilly did not arrive at this from a computer-science lab. He arrived from inside the slog. At West Point he audited chemistry and life-science classes, sat on a Critical and Creative Thinking Assessment committee, and coordinated studies for The Geneva Foundation. The pattern he noticed kept repeating: brilliant work in one field quietly answering a question being asked in another, with no search tool able to introduce the two.
Existing engines reward the keyword you already know. They punish the question you can't yet phrase. For interdisciplinary work - the kind that produces the most surprising results - that is exactly backwards.
Instead of a single search string, you feed it your own content. Instead of a ranked list, you get a map.
Drop in notes, a draft, or a whole project. Your document becomes the search, so you never have to guess the perfect query.
LitView reads across many concepts at once, pulling articles from public and private sources by relevance - not by keyword luck.
Results arrive as a 3D map of related ideas, so the connection between two distant fields becomes something you can actually see.
Latent Knowledge and Microsoft bonded over a shared faith in design thinking - building the tool around how researchers actually think. The figures Reilly's team points to when making that case:
Source: figures cited in Latent Knowledge's 2022 Microsoft partnership announcement.
Design-focused organizations, per the cited research.
Psychology and physiology. A master's in education. Triathlon start lines. Biotech consulting. Then a company. Reilly's resume reads like a man collecting vantage points - which is, more or less, the whole thesis of LitView.
A decade of triathlon teaches one thing above all: efficiency compounds. Shave a little wasted effort, repeat it a thousand times, win. He is doing the same thing to research workflows.
Trained in psychology and education before chemistry and biotech, he reads people and problems before he reads code - which is why LitView is built around how scholars think, not how machines index.
Board trustee of a soccer league, I-Corps entrepreneurial lead, consultant across biotech and GE Research. He keeps wandering into new rooms and noticing the doors between them.
When creating LitView, we were keen for it to be discipline-agnostic. Interdisciplinary students in particular struggle when it comes to undertaking research.- James Reilly
Reilly's stated goal is not a better keyword box. It is a world where the most useful paper - the one in a field you have never read - surfaces on its own, because the software understood the shape of your question better than your search terms ever could.