I keep a confession that would not surprise anyone who has run a lab for more than two years: the hardest part of research was never the thinking. It was finding the thing we had already written.
It started small. A reviewer asked for the exact wording of a comment from last year's proposal. A student needed a plot from a paper we published three cycles ago. Each request cost us an afternoon, and sometimes we never found the right version at all — so we quietly rebuilt something we already owned. The files were there. The knowledge was there. The path between the two had simply rotted.
If you nodded reading that, you are not lazy and you are not disorganized. The data says this is structural, not personal.
What the surveys keep telling us
A global survey commissioned by M-Files and fielded by Vanson Bourne put the pain in plain numbers: 96 percent of employees said they struggled to find the latest version of a document, and 83 percent had at some point recreated a file simply because they could not locate the one that already existed. Drop those percentages into a research group, where the same corpus is handed from one cohort of students to the next, and the waste compounds. New people re-index the same papers. Labs keep their own "close enough" reference libraries. Nobody means to — the knowledge base just forks and rots on its own.
What made it bite me personally was collaboration. A student switched laptops and the adviser's 300-megabyte proposal bounced off the email attachment ceiling. Someone writing a review opened three folders hunting a figure and found a version that did not match the one we had cited. The material existed. It was simply scattered, bloated, and unfindable.
Why our files swell the way they do
Most research material starts life as an Office document. A thesis draft might be only a few tens of thousands of words, yet the embedded illustrations, flowcharts, instrument screenshots, and scanned references often outweigh the text by a wide margin. Then nesting sets in: a chart in a slide deck copied from a spreadsheet, a diagram in a report cropped out of a PDF. Object inside object, the file swells, and nobody sits down to tidy it.
The result is that the thing we most need to move — a proposal, a manuscript, a set of supporting figures — is the thing least able to move. Cross-device sync stalls. Email rejects it. A reviewer asks for the source and we cannot even send it cleanly.
This is also where storage quietly bleeds. Komprise's customer research shows that 60 to 70 percent of data on enterprise network-attached storage has not been accessed in over 90 days, yet it still occupies the most expensive primary storage. Research institutions are the textbook case: a project closes, a paper publishes, and the raw data is touched by no one — but it cannot be deleted, because someday a reviewer will ask for it back, exactly as it was.
Universities face subject reviews and audits almost yearly. Pulling raw data from three or five years back is routine. By then nobody remembers which folder a final report landed in, and guessing by filename is the only option left.
The fix that finally felt real
For a long time the answer was "use a better search tool" or "just compress the PDFs." Those helped at the edges but never at the root. What shifted things for us was treating retrieval and compression as one problem, and keeping the whole thing inside our own network.
The retrieval side is a private knowledge base built on retrieval-augmented generation, deployed entirely on premise. The literature, the experiment logs, and the internal reports of faculty and students stay in the local machine room. They never cross into an external cloud. For a research organization, where data sovereignty is nearly non-negotiable, that is the floor, not a feature. You connect it directly to the databases the institution already runs, so the original data is retrieved and cited without being relocated. There is no need to start from scratch; connect and use. If your campus already runs a years-old literature system, it can stay and be reused rather than replaced.
The compression side is what made it portable. The compression engine we use recently reworked its entire Office pipeline so that embedded images and charts inside a document are compressed recursively, rather than only the body text. For papers and grant proposals, that is the difference that matters: the embedded visuals that were inflating the file finally get trimmed. The document shrinks, the email goes through, and sync across devices stops stalling. You can see the current product line behind this approach at https://www.uglypear.com/en/products/.
I want to be careful not to oversell a single release. But as a genuine recent improvement, recursive embedded-image compression is the kind of thing that changes daily behavior. Before, we would export a figure, re-save it, re-attach it, hoping to shave off enough to send the file. After, the document carries its own smaller weight, and the bottleneck moves elsewhere — which is exactly where you want the bottleneck to be.
The token economy nobody briefed us on
Here is the part I did not see coming. We now run literature synthesis, grant writing, and even routine coding through large language models, and those models bill by tokens. A figure pasted into a document as a high-resolution screenshot can quietly consume thousands of tokens every time it is fed to a model — even when the actual information in it is small. Compressing those embedded images before they ever reach a model does double duty: the file is lighter to store and to email, and the same content costs far less to reason over.
In published benchmarks, a 443-page OFD document finishes end to end in 124 seconds, and compressed images cut the token cost of feeding them to a model by roughly 85.9 percent, measured on a GPT-4o basis. For a lab that already pays for model access, that arithmetic adds up across hundreds of documents without anyone changing how they work. You do not need to adopt new habits. You need the file to arrive lighter. The full set of capabilities is outlined at https://www.uglypear.com/en/products/.
The ordinary risk is the real one
It is worth saying that leakage in our world rarely comes from a sophisticated attack. IBM's 2025 cost-of-a-breach study found that 26 percent of breaches trace back to human error — the more ordinary, more common path than external intrusion. In a research context that might be a collaborator's contact details left inside a shared draft, or a sensitive dataset attached to an email that went to the wrong address. The lesson is not to fear the spectacular; it is to make the ordinary case safe by default. With the private setup, sensitive content is detected and masked at ingest time, and retrieval respects permissions down to the fragment level, so a query returns only what a given person may see.
What it feels like in practice
During literature review, we pull the last five years of PDFs and notes into one place and ask in natural language to surface the relevant passages instead of paging through. When writing a proposal, last cycle's winning drafts, reviewer comments, and revision notes are all in the base, so a new section can borrow from how it was done before. For graduate training, advisers accumulate past opening reports and defense slides into a searchable experience base, so newcomers take fewer wrong turns. Across campuses, compressed documents are smaller and move faster, no longer blocked at the attachment ceiling.
These scenes sound unremarkable. The hard part is making them routine and undemanding. A knowledge base proves its worth precisely when nobody is tending it — yet it still finds and recalls.
I am not going to tell you our problems are solved. They are not, and a tool does not substitute for the discipline of naming files well. But the combination of keeping the data on campus, compressing the embedded weight out of our documents, and retrieving by meaning rather than by filename took a weight off that I had stopped noticing we were carrying. That, for a researcher, is a quiet kind of relief.
Try it now
If your lab or department is drowning in its own documents, start here:
- Browse the private RAG and document-compression products: https://www.uglypear.com/en/products/
- See the broader platform and how data stays on premise: https://www.uglypear.com/en/
- Read more deployment and selection thinking on the blog: https://www.uglypear.com/en/blog/
When material is under control, findable, and portable, research stops being slowed by its own documents.
FAQ
Q: How long does it take to get started?
A: There is no fixed period, because it depends on what you already run. The foundation connects directly to an institution's existing databases, so the original data is retrieved and cited without being relocated. Campuses with a current literature system or repository can often begin by connecting that source rather than building from zero, which keeps the early step small.
Q: Can it handle scanned PDFs and old theses?
A: Yes. Academic material frequently arrives as scanned references, instrument screenshots, and decades-old theses where the embedded visuals outweigh the text. The compression engine reaches embedded images and charts recursively, not just the body text, so a bloated scanned document or an old thesis draft can be trimmed and then indexed for content-based retrieval instead of filename search.
Q: Can a small lab with no IT staff use this?
A: Private deployment is designed to keep data inside your own network, and the operations burden is lower than building a retrieval system from scratch because it connects to databases you already operate. A lab without dedicated IT can start with a modest on-premise setup and grow; the point is to avoid standing up an entirely new infrastructure stack.
Q: How does it treat co-authored drafts from multiple people?
A: Retrieval respects permissions down to the fragment level, enforced at query time. That means a co-authored draft stays in one place, but each collaborator only sees the passages they are permitted to see, and sensitive content is detected and masked at ingest before it can surface in any query.
Q: Can students access it, or is it only for faculty?
A: Both. Faculty and students keep their literature, experiment logs, and internal reports in the same local knowledge base, and access is governed by the same fragment-level permissions. For graduate training, advisers can accumulate past opening reports and defense slides into a searchable experience base so newer students take fewer wrong turns.