Research in Intelligent Transport Systems
I lead the research and development efforts at Valerann, where I am responsible for overseeing a multidisciplinary team focused on building intelligent systems that empower road operators to manage their networks more efficiently through data and automation. Together, we developed a state-of-the-art AI platform that ingests and fuses data from a variety of sources to generate real-time insights about road conditions, safety events, and operational efficiency.
You can learn more about Valerann and our vision for the future of AI-enabled control centres at valerann.com.
Our Research Interests
My team and I are interested in a number of core research areas. This list is non-exhaustive due to the proprietary nature of some of our work. Of course, feel free to reach out if you want to learn more!
Multi-modal real-time data fusion — events, traffic, weather, risks. Our core approach is outlined in this https://patents.google.com/patent/US12118881B1/en?oq=US-12118881-B1.
Rigorous uncertainty quantification, as well as understanding the role each data source plays as part of a wider data fusion matrix. We have published a few papers around this and have presented at TRB.
Fuzzy logic and traceability — we have developed an in-house natural language rule engine to help road operator customise our fusion engine with ease and precise control.
Computer vision as applied to moveable road side cameras — calibration, geo-referencing, event detection (stopped vehicles, hazard lights). We patented a lot of technologues around this, for example this work on automated geo-referencing.
Traffic modelling and anomaly detection. Lately, we have been quite interested in the concept of foundational traffic model and end-to-end models that can understand traffic patterns.
ML driven incident risk modelling and using it for data fusion prior.
Agentic Control Center is a concept that we are actively exploring, especially how the latest wave of LLM technology can be incoporated into a control room to enhance event validation.
From Research to Solution
Beyond the research, I was also responsible for the productionalisation and large-scale deployment of our algorithms in cloud environments. This included ensuring scalability, low-latency inference, and maintainability of the models in production. One of the most significant challenges—and a core area of our interest—is the transferability of models across different geographies, where infrastructure, driving behaviors, and incident semantics vary significantly. Addressing this involved a combination of domain adaptation, retraining pipelines, and dynamic model tuning. We also invest a lot of effort in monitoring, observability and visualisation.
Leading Innovation Activities
Innovation was not limited to product development. I actively lead initiatives in the organisation such as:
Authoring patents and protecting IP
Organizing and participating in internal and external hackathons
Representing Valerann in academic and industry conferences
Facilitating Journal Clubs to maintain cutting-edge knowledge and foster team learning