Our team brings research experience from the University of Toronto Machine Learning group and Vector Institute. We develop small-model reasoning, agentic systems, and Chinese-language NLP as deployable, auditable systems.
Jointly optimising reasoning and self-refinement under one binary reward — no critic model, no process reward, no critique data. Adds +11.5 pp on AIME pass@4 (Qwen3-4B) over vanilla GRPO.
Developed with Xiaohongshu, this mobile coding-agent benchmark includes 50 production iOS feature tasks, PRDs, Figma specifications, and 449 hand-written tests; the best-performing evaluated agent-model pair completes 12%.
Coordinates documents, search, databases, and business tools, decomposing complex work into parallel, traceable workflows.
A multi-stage RL system for pre-publication correction across text, subtitles, transcripts, and OCR, tuned for false-alarm control and deployment efficiency.
Detects spoofed sites and malicious redirects, with knowledge-graph evidence, risk classification, and remediation guidance.
Covers 50+ attack patterns with on-premises deployment and continuous monitoring for models and agentic systems.
Converts financial reports, contracts, and invoices into structured data, including complex headers, merged cells, and low-resolution inputs.
Coolwei AI Lab is an innovative team focused on cutting-edge research and applications in artificial intelligence. Our mission is to advance AI technology and create value for enterprises and society.