Applied AI
Web3 Analytics Dashboard
An on-chain dashboard from CSV exports (Dune or otherwise) — KPIs, series, cohorts. Blockchain as a data terrain, not the Lab’s identity.
- Problem
- On-chain exports are readable in SQL, rarely as an analysis product. Missing piece: KPI + retention views on a standard CSV.
- Why it matters
- Applied AI / data. It shows the Data thread on an adjacent domain without turning the site into a crypto portfolio.
Stack: Python · Streamlit · Plotly · pandas
Architecture
Streamlit app: CSV upload → detect date / address / value columns → KPIs → Plotly charts (volume, tx count, top addresses, retention heatmap).
Optional Dune API. Target deploy: Streamlit Cloud.
Data
On-chain analytics CSV exports. Not a node indexer. Quality follows the export, not a documented custom collector.
Approach
- Flexible schema: the user maps columns.
- Explicit KPIs (volume, tx, unique addresses, active days, averages).
- Interactive charts rather than an ML model.
Results
Public repo, Streamlit UI, views listed in the README. No accuracy benchmark — this is an exploration cockpit, not a detector.
Limitations
No Ethereum node, no original on-chain score, no “production trading” claim. Crypto aesthetic is intentional: the Lab files it as Applied AI, not a flagship.
Learnings
An adjacent domain (blockchain) only enters when it serves a Core method: here, wrangling, KPIs and time series. Older on-chain ML (anomalie_eth, MLflow) stays learning evidence, not a Lab product.