Tania
Uncover hidden null results with AI insights.
Created on 6th June 2025
•
Tania
Uncover hidden null results with AI insights.
The problem Tania solves
Scientific research has a hidden problem: null result bias. This happens when researchers run experiments that don’t support their original hypothesis, and those results never get published. As a result, other scientists might unknowingly repeat the same failed studies, wasting valuable time, funding, and effort. It also creates a distorted view of what we know, because only the successful or positive outcomes are visible in the literature.
Tania was built to help solve this. It’s a research assistant that uses natural language processing and knowledge graphs to uncover potential null results that are buried in the text of published papers. It looks for subtle language that suggests something didn’t go as planned, and maps how scientific concepts connect to spot areas where hypotheses seem to have quietly failed.
Tania makes life easier for researchers, meta-analysts, and science funders. It helps them avoid redundant experiments, identify weak or inconclusive areas in research, and make more informed decisions. And by storing key insights on the Solana blockchain, Tania also ensures that this information is accessible and tamper-proof for the long haul.
Challenges we ran into
One of the challenges we faced early in the project was a technical mismatch between our preferred tools and the structure of the hackathon starter guide. The starter materials provided for the BIOHackathon were built using JavaScript and TypeScript. While this made sense for plugin development, our team planned to implement the core NLP pipeline using Python, which offers more robust support for machine learning and text analysis tasks.
This led to a key architectural decision: we chose to separate the plugin logic from the NLP engine. The idea was to have the JavaScript-based plugin communicate with our Python NLP service through HTTP requests. However, we were initially unsure whether this hybrid approach would be acceptable under the hackathon rules.
To resolve this, we reached out to the hackathon moderators to seek clarification. They quickly got back to us with approval, confirming that our approach was valid. This gave us the confidence to move forward without compromising on our technical strategy, and it was encouraging to feel supported by the organizing team so early in the build process.