
Three AI systems.
Built solo. Live in production.
Content operations, investment research, and enterprise document generation — each one designed, built, deployed, and operated end-to-end. Not prototypes. Not slideware.
content-os
Point it at a topic and it gathers grounded sources, drafts platform-native variants, and queues them for a human approval gate — publishing to six channels the moment you say go.
- Grounded news harvestingPulls and scores live sources, keeps citations attached — no ungrounded generation.
- Six-channel variant generationOne draft becomes native copy for LinkedIn, X, Discord, Telegram, WhatsApp, and Instagram in one pass.
- Titan agent orchestratorDescribe a goal in plain language; a planner decomposes it across researcher, editor, and formatter agents.
- No timeout cliffRedis-backed job queue turns 4-minute generations into a fire-and-poll flow instead of a blocking request.
AI Investment OS
A multi-factor signal engine — technical, fundamental, sentiment, and market regime — scores a watchlist daily, with a full risk and backtesting suite behind every call.
Paper research mode. Not investment advice.
- Ensemble signal scoringTechnical + fundamental + AI news sentiment + regime modifier, with confidence and risk on every symbol.
- Full risk suiteVaR, Sharpe, Sortino, Calmar, drawdown, and DCF/DDM valuation, not just a buy/sell badge.
- Monte Carlo + walk-forward backtestsStress-test a strategy against thousands of simulated paths before it ever sees real capital.
- Real brokerage, real guardrailsLive Snaptrade sync for positions and balances; order placement is admin-gated and paper trading is the default.
AI Document Workbench
Ingest source-of-truth documents from SharePoint, Confluence, or S3, and a multi-agent pipeline drafts, reviews, and scores technical documents against your own gold-standard templates.
- Multi-source ingestionSharePoint, Confluence, S3, or local files — extracted into a searchable knowledge base automatically.
- Document-derived knowledge graphEntities and relationships extracted from your documents, with full lineage tracing back to source.
- Multi-agent generation pipelineGenerator, reviewer, improver, and researcher roles collaborate on every draft, not a single prompt.
- AI quality scoring that self-correctsTen weighted dimensions score every draft; the pipeline keeps improving until it clears your bar.
How it's actually built