// 01 > CONTEXT
Felis AI was built for Vanshika Tyagi, a research scholar who has to read through a large volume of research papers. The goal was a tool that lets a scholar hand their papers to an application and query across them in plain language, instead of re-reading everything by hand.
// 02 > THE_PROBLEM
Research scholars read a huge volume of papers, and finding the passage that answers a question means combing back through all of them by hand.
A scholar's working set is dozens of dense papers. Keyword search misses paraphrased ideas, and there was no way to ask a question across the whole corpus and get a grounded answer that points back to the source.
Before
Dozens of papers to re-read · keyword search misses meaning · no cross-paper querying
// 03 > APPROACH
I built a RAG application where a scholar uploads their papers and queries across them in natural language, with retrieval grounding every answer in the actual source and persistent memory across a research session.
Upload and query papers
Scholars add their own research papers; the app retrieves the relevant passages and answers in plain language.
Grounded, with memory
Every answer is grounded in the retrieved source, and persistent memory carries context across a long research session.
// 04 > THE_RESULT
A research scholar can hand Felis her papers and ask questions across all of them, instead of re-reading by hand. Private engagement.
Before
- Re-reading dozens of papers
- Keyword search misses meaning
- No querying across the corpus
After
- Upload papers, ask in plain language
- Answers grounded in the source
- Memory across the research session
“I read a huge number of research papers, and Felis changed how I work. I hand it my papers and ask questions across all of them instead of digging back through each one, and it points me to the right passage.”