TORCS Corkscrew Revisited: Training a SAC Agent From Scratch With Cleaner Reward Engineering
TORCS Corkscrew Revisited: Training a SAC Agent From Scratch With Cleaner Reward Engineering
Featured Projects
arXiv RAG System
Hit Rate 60→100%
MRR 0.51→0.82
async pipeline
End-to-end RAG for academic papers, built and optimised from scratch.
FastAPI
lifespan startup, fully async I/O, and a documented
28pp fine-tuning regression traced to training-data contamination.
FastAPI · ChromaDB · LLM (Qwen3 4B / Ollama) · Docker
FinScope — Multi-Agent Analyst
3-agent LangGraph
~15s parallel
hallucination check
Multi-agent RAG over SEC EDGAR & Companies House filings.
Retriever → Analyzer → Critic pipeline, parallel analysis via
asyncio.gather, and a Critic Agent with a conditional retry loop.
LangGraph · Groq (llama-3.3-70b) · ChromaDB · FastAPI · Langfuse
DefectVision
100% AUROC (bottle)
normal-only training
real-time webcam
Real-time manufacturing defect detection with PatchCore — trained on
normal images only, no labelled defects. FastAPI inference API plus a
Streamlit dashboard with live webcam streaming.
Anomalib · PatchCore · PyTorch · OpenVINO · FastAPI · Streamlit
TORCS RL Racing Agent
Recent Posts
Reply Mirror AI Challenge 2026: Multi-Agent Fraud Detection Under a 6-Hour Clock
Reply Mirror AI Challenge 2026: Multi-Agent Fraud Detection Under a 6-Hour Clock
LLM-as-Judge for Hallucination Detection: Does the Critic Agent Actually Work?
This is a follow-up to From arXiv to SEC: Building a Multi-Agent Financial Report Analyst with LangGraph. That post ended with: “The remaining question is...
DefectVision: Building a Real-Time Manufacturing Defect Detector Trained on Normal Images Only
DefectVision: Building a Real-Time Manufacturing Defect Detector Trained on Normal Images Only
From arXiv to SEC: Building a Multi-Agent Financial Report Analyst with LangGraph
From arXiv to SEC: Building a Multi-Agent Financial Report Analyst with LangGraph