Kallisto-1: Quantitative Execution Engine

Autonomous quantitative trading engine with VIX macro gating, Tri-Agent consensus, and Alpaca atomic bracket routing to insulate momentum portfolios.

Comment

Inspiration

What it does

How we built it

Challenges ## Inspiration

Momentum strategies systematically fail during macroeconomic regime shifts. In severe market volatility spikes—such as the 2022 market downturn where benchmark indices fell over 34%—standard trend-following algorithms suffer catastrophic drawdown due to false breakout traps and adverse spread slippage. We engineered Kallisto-1 to solve this structural vulnerability by creating an autonomous execution engine that insulates momentum portfolios using macroeconomic volatility gating and sub-3ms Tri-Agent Quorum consensus.

What it does

Kallisto-1 operates as an institutional-grade quantitative execution engine:

  • Macro Volatility Gating: Continuously evaluates VIX term structure dynamics (contango vs. backwardation) to suppress low-conviction breakout trades during market stress.
  • Tri-Agent Quorum Consensus: Reaches unanimous trade consensus across three specialized pillars in sub-3ms:
    1. Macro Intelligence (NVIDIA Nemotron-3.5): Evaluates systemic regime stability.
    2. Microstructure Analysis (L2 OFI): Measures Level-2 Order Flow Imbalance and Wyckoff volume absorption.
    3. Tail-Risk Boundary (CVaR): Enforces strict real-time Conditional Value-at-Risk limits.
  • Atomic Convex Bracket Routing: Employs the Alpaca Trading API to dispatch atomic bracket orders (Take-Profit & Stop-Loss) locking in a 2.5:1 asymmetric reward-to-risk ratio while bounding maximum adverse excursion at -1.8%.
  • Enterprise Governance: Generates a real-time cryptographic JSON audit trail with microsecond precision, backed by an interactive emergency Kill Switch that instantly collapses exposure to a 100% cash shield.

How we built it

  • Engine Core: Python 3.11 and FastAPI for high-throughput, deterministic signal processing and risk bounds enforcement.
  • Broker Gateway: Alpaca Trading API v2 for live atomic bracket order routing.
  • Visual Telemetry: Next.js and Tailwind CSS delivering real-time L2 order flow inspection, telemetry, and interactive circuit breakers.
  • Empirical Backtesting & Verification: Backtested across 1,042 trading days net of 15 bps friction, verified using Playwright headless browser tour automation.

Challenges we ran into

  • Low-Latency Consensus: Synchronizing decisions across macro reasoning, L2 order flow imbalance, and CVaR calculations within tight latency budgets required an asynchronous execution pipeline.
  • Bracket Order Synchronization: Guaranteeing atomic execution on Alpaca without fill slippage during sudden volatility spikes required rigorous pre-dispatch validation.

Accomplishments that we're proud of

  • Slashed maximum drawdown from -34.83% (naive benchmark) down to -11.89% across 1,042 trading days.
  • Expanded the Sharpe ratio from 0.94 to 2.14 and achieved a Calmar ratio of 1.84.
  • Deployed a fully containerized, responsive production terminal live on Vercel with zero runtime drift.

What we learned

  • Quantitative risk management is most effective when multi-tiered: global macroeconomic trend gating must strictly constrain local microstructure order flow signals.
  • Structuring trades with 2.5:1 convex payoffs and rigid downside tail boundaries creates resilient compounding across distinct market regimes.

What's next for Kallisto-1

  • WebSocket-based high-frequency tick data processing.
  • Multi-asset execution spanning currency pairs and fixed-income ETFs.
  • Adaptive machine-learning bracket sizing driven by live implied volatility surfaces.we ran into

Accomplishments that we're proud of

What we learned

What's next for Kallisto-1: Quantitative Execution Engine

Built With

  • alpaca-api
  • docker
  • fastapi
  • next.js
  • nvidia-nemotron
  • playwright
  • python
  • tailwindcss
  • vercel
Share this project:

Updates

Submission history