Transforming audited financial statements
into structured business intelligence.
01 — Problem
Every 10-K, 10-Q, and earnings transcript lands in its own format, and the line that actually matters is usually a footnote, not the headline number. An equity team can lose the first two days of earnings season just reading, before any real analysis starts.
02 — Solution
The agent parses each filing's XBRL alongside its full narrative text, diffs it against the prior period, and drafts a brief where every claim carries a citation back to its exact source line — nothing reaches an analyst unsourced.
03 — Extraction Pipeline
Filing Intake
10-K · 10-Q · Transcripts
Structured Parse
XBRL
Narrative Extraction
Chunk + Embed
Retrieval Index
Vector DB
04 — Architecture
Analyst Console
Next.js
Retrieval + Drafting
Claude
Citation Verification
rejects unsourced claims
Historical State
Postgres
05 — Engineering Decisions
06 — Business Impact
0
Reports processed
0
Analyst time saved
0
Insight accuracy
07 — Future Scope
Live for citation-verified briefs across 300+ tracked tickers, expanded from 60 without adding headcount.
Cross-filing pattern detection to flag unusual language shifts across a company's last several quarters automatically.
Resources
01 — Documentation
02 — Source
Private source code.
Available during technical discussions.
Tech Stack
Frontend
Next.js
Backend
Next.js API Routes
AI
Claude
Finance
XBRL · Financial Data Parsers
Infrastructure
PostgreSQL · Vector DB
Interested in this project?
Let's discuss it — schedule a live walkthrough.