9× faster inference
Real-time visual SLAM on edge hardware
A state-of-the-art SLAM pipeline taken from 0.26 to 2.4 FPS on a Jetson Orin NX — a ~9× speedup under strict power and compute limits.
Selected Work
+20% F1 score
Defect detection where models plateau
A 20% F1 improvement on production inspection data defined by tiny targets, severe class imbalance, and almost no labelled failures.
Auditable by design
Agentic generation in a regulated domain
Health-protocol generation with an LLM-as-a-judge review layer, built so that every output can be audited rather than trusted.
Shipped to production
Hybrid multimodal retrieval for visual collections
Search across a heterogeneous image collection by fusing vector-indexed image and text embeddings with keyword search.
−40% analysis time
LLM survey analysis at research scale
An LLM/RAG platform for user-research teams that cut analysis time per project by 40%, measured against human-coded baselines.
750+ stars, 200+ forks
tf-cnnvis — seeing inside a convolutional network
An open-source CNN visualisation library for TensorFlow, now past 750 stars and 200 forks.