An experimental research pipeline: public market data becomes wallet profiles, clusters, heuristic signals, and optional XGBoost scores. It does not place live trades.
Implemented market and trade ingestion, wallet analysis, and clustering.
Built freshness, funding, timing, and activity signal detectors.
Added feature extraction and ML scoring. The backtest engine and terminal dashboard remain design specifications.
A Poke-inspired assistant with an interaction agent, execution agents, Gmail tools, and a scheduler. The repository ships a Next.js web interface; the WhatsApp scene here is a conversation concept.
Separated conversation from background execution in a multi-agent backend.
Connected Gmail actions through Composio and added reminder scheduling.
Built a web UI for the assistant. The illustrative phone preview describes the interaction, rather than a released iOS client.