Applied Labs

Applied AI systems as a foundation for research.

The labs connect Evening Star research with working systems across vulnerability intelligence, LLM security, software assurance, and decision support. They support technical writing, evaluation, and investigation of how AI behaves in practice.

This page shares research projects and their methods. It does not offer consulting, freelance development, or client implementation services.

Research systems and publication access

Purple Radar Applied research lab

Evidence-first vulnerability analysis for modern attack surfaces.

Purple Radar applies Evening Star thinking to vulnerability records, exploit signals, exposed services, scanner confidence, known-exploited indicators, and risk-prioritized findings.

CVE correlationExploitabilityRisk scoring

Read the Purple Radar research

Purple Firefish AI security research lab

Adversarial input detection before prompts touch the model.

Purple Firefish is the gateway system for prompt injection, jailbreak behavior, data-exfiltration attempts, delimiter abuse, obfuscation, and system-prompt probing signals.

Prompt injectionJailbreaksLLM defense

Read the Purple Firefish research Explore the external Firefish project

Candles Edge Market analytics research

AI-assisted market analytics for reading noisy signals.

Candles Edge inspires Evening Star's market adapter: OHLCV features, returns, ATR percentage, Bollinger width, RSI, MACD, volume z-score, anomaly markers, and decision support.

Anomaly markersRegime signalsDecision score

Read the Candles Edge research Explore the external Candles Edge project

Beacon Research Platform

Governed research access for compatible AI clients.

A governed MCP gateway connecting compatible AI assistants and agents to the Evening Star research catalog. Read-only Research access does not enable private lab or administrative capabilities.

MCPResearch AccessAI Governance

Connect to Beacon

Lab Philosophy

Each lab tests the same pattern in a different domain.

The lab systems differ by domain, but our pattern is consistent: observe context, detect meaningful change, explain uncertainty, and evaluate how model output can support human decisions.

External project links provide context for the research. They are not invitations to commission services through Evening Star AI.