EPulse Event Discovery.
Engineering case study by Priyansh Dobariya · Python · Hermes Agent · Gemini · Google Sheets
Find events worth reviewing.
An AI-assisted event-research pipeline that extracts conference opportunities, checks duplicates, and saves sourced records for human review.
Behind the build.
Query generation and extraction
Validation and deduplication
Source URLs and confidence for review
Tools & technologies
What this solves
Finding relevant future trade shows and niche conferences takes repeated searching and manual record keeping. The first phase narrows that research into a reviewable event list.
How it works
Python services generate category-based searches, collect public page text, ask Gemini for structured event fields, validate records, and check likely duplicates. A Google Sheets repository stores the candidate list with source URLs and confidence levels; Hermes Agent orchestrates the discovery workflow.
From input to outcome
A category produces search queries. Candidate pages are extracted into structured event data, checked for missing or conflicting fields, compared with existing rows, and saved as new or updated review records.
Decisions visible in the code
- Source URLs and confidence travel with each event record so a person can verify dates and relevance.
- A Sheets store makes the Phase 1 output easy for operators to inspect and correct.
What the workflow enables
Researchers receive a consolidated list of candidate events with evidence attached instead of redoing the same searches.
Phase 1 stops at discovery. It does not approve events, book travel, make payments, or contact organizers.