RegWatch

Compliance teams at companies operating across multiple jurisdictions spend hours manually tracking regulatory feeds across SEC, FCA, ESMA, SEBI, GDPR and others. Missing an update or misreading an amendment can mean fines in the millions. RegWatch automates this entirely. A company inputs their industry and operating regions. RegWatch fans out parallel web fetch agents across all relevant regulatory sources, aggregates and deduplicates results, filters for company-specific relevance using sentence-transformers embeddings, and runs a violation checker that classifies each item as VIOLATION, AT_RISK, WATCH, or CLEAR with a confidence score and clause reference. A dedicated Jurisdiction Conflict Detector catches cases where complying with one region's rules creates liability in another — GDPR's right to erasure conflicting with SEC's seven-year record retention requirement being a classic example. These are surfaced as separate CONFLICT flags with structured resolution recommendations. Every high-severity flag gets a generated remediation memo with actionable steps, recommended owners, deadlines, and a precedent note pulled from a FAISS-indexed dataset of real historical enforcement actions. A compliance score from 0 to 100 tracks the company's overall posture across runs, and a risk fingerprint radar chart shows exposure across six regulatory domains. The pipeline is built on LangGraph with parallel node execution via the Send() API, with Mistral 7B Instruct served through vLLM 0.17.1 on AMD Instinct MI300X via ROCm 7.2.0. The parallel violation checker calls are batched as concurrent async inference requests, directly leveraging AMD's memory bandwidth for throughput. The backend runs on FastAPI and the frontend on React with TailwindCSS.