Aegis Whistle

AegisWhistle—the world’s first Voice AI Ethics Officer, built on Llama 3 (Maverick-17B-128E-Instruct) and other OSS tools We’re pushing the boundaries of ethical AI governance with: Technical Stack & Innovation Voice-AI for Accessibility: VAPI.ai + Tavus API for low-latency, multilingual voice+ video interactions (whistleblowers speak naturally; no typing required). Google AI Studio for sentiment analysis to triage high-risk reports. Llama 3 Maverick Fine-Tuning: Custom-training on 50K+ legal/ethics documents (SEC, GDPR, etc.) for precise policy gap analysis. Structured JSON outputs via tool calling to auto-generate audit trails. Transparency Backbone: Hyperledger immutably logs reports, while GROQ’s LPUs enable real-time RAG over compliance databases. ML Techniques Deployed Technique: Use Case RAG: Pull real-time regulatory updates (e.g., EU AI Act). Agents: Multi-step workflows (e.g., interview whistleblowers → escalate to HR). Fine-Tuning: Adapt Llama 3 for legal/ethics reasoning. Structured Outputs: Generate compliance reports in JSON for APIs. Voice Models: Convert speech → actionable insights.