Implemented · paper trading by default
Multi-Agent Capital Pipeline
Built an Indian equities research engine with eight parallel agents, including an agent for geopolitical signals.
- Engineering impact
- Connected research, risk decisions, simulated execution and performance review in one workflow.
- Business impact
- The system supports strategy evaluation with paper trading as the default.
Problem
Fundamentals, liquidity, governance and geopolitical events can point in different directions. A useful workflow has to preserve the evidence behind a recommendation and constrain portfolio exposure.
My role and ownership
Built the research pipeline that combines independent signals, applies portfolio risk rules and exposes position and performance data in a dashboard.
Architecture and tradeoffs
Python runs eight analyses concurrently and collects a common result with confidence and risk flags. An aggregation stage checks agreement, then a market-regime stage adjusts sizing. Portfolio constraints sit between signal generation and the broker adapter. FastAPI serves the dashboard.
Workflow in use
Scan the configured universe, review signals and risk flags, simulate orders, and review open positions and performance on the dashboard.
Scope and constraints
This is the Multi-Agent Capital Pipeline, a separate project from JFT. Realized performance after costs has not been verified for publication.