Hi HN! I'm George, founder of ValueSense (valuesense.io). For the past year, I’ve been working on a platform that helps investors research public companies the way analysts at hedge funds do, but without the spreadsheets, PDFs and 12 browser tabs.

The backstory I used to rely on Dataroma to track superinvestor portfolios (the Buffetts, Ackmans, Klarmans of the world). I liked the simplicity, but as I used it more, I hit major friction: - There’s no conviction scoring, trends or clustering of buys/sells - No institutional or insider context - No real ability to explore investor relations data - UI hasn't changed in over a decade

I actually tried to acquire Dataroma at one point, but the deal didn’t go anywhere. So I started building the tool I wanted.

What I’m building The core idea: a research engine that connects smart money activity with investor relations data and makes it usable.

Here’s what’s already working: 1) Smart Money Signals A pipeline that ingests, cleans and structures data from: - 13F filings — fund-level holdings, position sizes, entry timing - Insider trades — Form 4s parsed for clusters, trends, and volume - Institutional flows — sourced from ownership filings (13G, 13D, NPORT, etc.)

We generate: - Conviction scores — based on % of portfolio, position history, and co-investing behavior - Cluster flags — when multiple insiders or funds pile into a stock at once - Time-series of ownership shifts — visualized by entity and role (e.g. activist, insider, fund)

This is all stored in a PostgreSQL database with event-based indexing and rendered live with a charting engine that uses caching for fast reloads across tickers.

2) IR Intelligence

The other side of public company research is buried in PDFs: earnings decks, segment data, KPIs, commentary etc.

I built a parser that pulls these into a structured format:

- Revenue by segment and geography - Operational KPIs (e.g. Uber trips, Netflix users, Nvidia DC revenue) - Historical earnings slides and management guidance - How Company Makes Money breakdowns

It runs on a data pipeline built in Python + Airflow, pulling from SEC EDGAR, earnings call transcripts and investor websites. All numbers are standardized quarterly and TTM, cleaned, and visualized inside the platform.

My Technical stack Backend: FastAPI, PostgreSQL, Redis ETL: Python, Airflow, BeautifulSoup, custom EDGAR parser Data storage: Postgres for structured financial data; S3 for raw filings & charts Frontend: React + Tailwind, Highcharts for data visualization Infra: GCP + Cloud Run + Supabase auth AI: LLMs used for DCF templating, narrative parsing, and user-defined screeners

What I’m working on next:

Letting users ask questions like: “Which stocks have both rising insider buys and top-line revenue growth?” “What did Ackman add last quarter that others didn’t?” This is a combo of natural language → SQL generation and curated filters. DCF and valuation models that users can tweak, save, and share AI research agents trained on historical investor letters, filings, and segment data

I’m not launching publicly yet — just shipping core modules and talking to early users. But I wanted to share here because:

- A lot of folks on HN manage personal portfolios and feel the same frustration - Many financial tools today are either too surface-level (Yahoo Finance) or too expensive (Bloomberg)

If anyone’s built similar data pipelines, financial tools, or research systems — I’d love to trade notes

Also, if you’re building a fintech product or have thoughts on data infrastructure, LLMs for research, or public markets — I’d love to hear how you’d make this better.

Try it (early version): https://valuesense.io Email me: george@valuesense.io

Happy to answer anything!