Developing a context-aware, multimodal Small Language Model (SLM) interface for real-time, privacy-preserving decision support in high-risk scenarios. Leading the design and deployment of an explainable Human-Robot Interaction (HRI) system to improve operator task accuracy and reduce cognitive load during critical missions. Leads a team of five researchers and engineers and contributes to competitive research grant proposals.

Building an enterprise AI engine for workflow automation: RAG-based retrieval over vector databases to ground LLM responses in private organisational data, a hybrid multimodal-LLM framework supporting both local and remote models, and scalable pipelines for vector synchronisation and semantic search.

Developed and maintained large-scale video deduplication services supporting real-time content moderation on TikTok LIVE. Designed and deployed computer vision and multimodal models for duplicate and near-duplicate detection across visual, audio, and textual signals, and built scalable inference and retrieval pipelines over high-volume livestream data. This work is published as two KDD 2026 papers.

Research under the iGyro (Information Gyroscope) project on detecting and assessing AI-generated and manipulated visual media. Developed algorithms for analysing deepfakes and AIGC content, and led dataset construction, experimental evaluation, and manuscript preparation for peer-reviewed venues.

Interdisciplinary project Deep and Cheap-Fakes — Effects on Audience’s Attitudes, Knowledge, and Literacy, joint between computer science and communications research. Produced large-scale curated deepfake datasets used as controlled stimuli in audience studies on media perception and literacy, underpinning four journal articles in communication and behavioural science venues.

Contributed to a TikTok video duplication detection project using transformer-based model variants and image-matching algorithms for large-scale video similarity analysis.

Computer vision solutions for automated safety-violation detection on construction sites: object detection, multi-object tracking, human action recognition, pose estimation, and proximity analysis. Contributed to end-to-end deployment of vision-based monitoring systems in real-world industrial environments.

Full teaching, curriculum design, and final-year project supervision load. Supervised three final-year project groups, guiding one to a peer-reviewed publication.

Core teaching across theory and programming modules; contributed to departmental curriculum development and undergraduate academic mentoring. Initiated a bioinformatics/machine-learning collaboration that produced four peer-reviewed journal and conference papers.

Thesis on Digital Face Synthesis: Safeguarding Against Threats. Published papers in ICCV, WACV, IJCB.