ml-portfolio-suite
Five live ML demos: Kalman, CV, bandits, CLIP, quantization
Explore how defenses cut AI agent scams in a simulated economy
Control AI compute per query to save cost and keep accuracy
Detect and halt risky AI actions using internal safety sensors
Record and pinpoint the first mistake in AI agent runs
Explore early detection of AI reward hacking
Explore how defenses cut AI agent scams in a simulated economy
Control AI compute per query to save cost and keep accuracy
Detect and halt risky AI actions using internal safety sensors
Record and pinpoint the first mistake in AI agent runs
Explore early detection of AI reward hacking
Generate→Filter→FineTune loop with quality pipeline
Speculative decoding with rejection sampling, 1.87x faster
GPT-4o structured JSON extraction with validation layer
Multi-hop QA benchmark showing where flat RAG fails