Active work on agent orchestration and AI business automation, alongside the research systems and sensing platforms built across my PhD and postdoc.
My own autonomous agent orchestration stack built on top of OpenClaw, Hermes, OpenGOAL, and MCP (Model Context Protocol). Reusable patterns for multi-agent coordination, tool use, and DAG-style workflow execution — the technical backbone for the SME automation platform.
AI-driven workflow automation for Australian SMEs — replacing labour-intensive back-office processes (claims, document review, customer triage, compliance reporting) with autonomous agents. Productisation of the orchestration stack into measurable customer ROI.
Large-model-assisted phone call enhancement using a smartphone's built-in accelerometer. First-author work integrating LLMs with sensor data for real-time speech enhancement. Accepted at IEEE INFOCOM 2026 (CORE A*).
End-to-end AI platform for a Fortune 500 retail client: phone-sensor-only indoor positioning across >10,000 m² stores, no cameras, no human-in-the-loop. Composed of autonomous agents (sensor fusion, posture classification, trajectory prediction) shipped via Docker + CI/CD.
Greenfield AI development for a Fortune 500 client's smartwatch gesture recognition. Replaced a failing classical pipeline with deep learning — 95%+ accuracy across 8 gesture classes, 50%+ inference-latency reduction through model compression for real-time on-device deployment.
Privacy-preserving facial expression recognition using millimetre-wave radar. No camera required — captures subtle facial muscle movements from RF reflections. Published at MobiCom 2023 (CORE A*).
Contact-free blood pressure monitoring using commodity mmWave radar. Estimates systolic and diastolic BP from pulse wave velocity measured via radar reflections. Published at SenSys 2022 (CORE A*).
On-device deep learning for mobile devices: ChainSGD-Reduce for distributed training across a chain of smartphones, plus a release-and-inhibit control mechanism for resource-efficient inference. Published at IEEE TMC, IEEE/ACM ToN, SenSys, and IPSN.
Real-time phone call quality enhancement using the smartphone's built-in accelerometer. Detects call conditions and adapts audio processing without extra hardware. Published at UbiComp 2025 (CORE A*).