FacultyPulse
Finds new papers, awards, and news for 40 robotics faculty, and filters out the wrong matches before anyone posts them.
- ~5 min
- Weekly research, down from ~5 hours
- 40
- Faculty tracked every scan
- 0 → 500
- LinkedIn followers in six months
Northeastern's Institute for Experiential Robotics posts about its faculty's work on LinkedIn, and finding that work meant searching for each of 40 people by hand every week. FacultyPulse does the searching, and a person reviews what it finds before anything goes out.
How it works
Each scan runs four targeted searches per faculty member through Tavily, which reaches arXiv, IEEE, ACM, journals, and news sites. An LLM (Claude or OpenAI, switchable) then rejects the usual false hits: a different person with the same name, an old page, or a faculty profile. What survives lands in a React and TypeScript dashboard on a Flask API, where it can be filtered, reviewed, and exported.
Picking a search source
Tavily charges per query, so I added Google News as a free option and a script that runs both over the same faculty and compares what each finds. Google News missed publications entirely, and those are the largest category, so Tavily stayed the default and Google News became the cheaper choice for news and awards.
What came of it
Weekly research went from about five hours to about five minutes. The posts it fed helped grow the institute's LinkedIn page from zero to about 500 followers in six months, and the team kept using the tool after I documented it and handed it over.