Research

“I know my 💩,” is a badge ribbon I humorously picked up at a recent conference. Ironically, I argue the exact opposite. Wastewater has repeatedly reminded me of how little we know about its ever-evolving nature and the complexity of its treatment processes. Since 2015, my curiosity-induced quest has been to research and develop bioprocesses for a resilient and smart urban water infrastructure.

My research is anchored at the intersection of wastewater process engineering, environmental biotechnology, and data science. A significant part of our work focuses on mechanistically investigating how microbial communities in bioprocesses respond to operational and environmental conditions, with the goal of kinetically tailoring these communities to develop new technologies and/or enhance the performance of existing processes. While such mechanistic insights are critical for foundational understanding and process design, they sometimes fall short in capturing the dynamic complexity of full-scale wastewater treatment systems. To bridge this gap, we integrate machine learning (ML) into our research, embedding domain knowledge and leveraging emerging explainability tools. This allows us not only to build highly predictive models for full-scale processes but also to uncover underlying drivers and provide meaningful decision-support tools.

Research Areas

High-rate activated sludge system (HRAS) for resource recovery from wastewater

HRAS is characterized by operating at high substrate loading rates, which trigger a non-oxidative microbial response instead of oxidation. As a result, higher fractions of incoming wastewater resources are diverted toward recovery streams, upstream of the endo-energetic oxidative processes. Nonetheless, the present understanding of HRAS bioprocesses remains limited, and the existing literature is fragmented and often contradictory. This uncertainty has led to suboptimal operation of such bioprocesses and missed opportunities for enhanced resource recovery, which we aim to address in my research.

Continuous flow densification of the activated sludge process

Continuous flow (CF) densification is an emerging strategy that transforms conventional activated sludge into densified activated sludge (DAS) by encouraging the formation and retention of densified biological flocs (DBF), alongside ordinary flocs (OBF), while selectively washing out poorly settling pin flocs and filamentous bacteria. This approach has shown promising benefits at wastewater treatment plants (WWTPs), particularly in improving sludge settleability and process efficiency. However, its outcomes remain inconsistent across facilities, with key uncertainties around the extent of achievable densification, its stability over time, and the operational strategies needed to maintain it. Our research lab addresses these uncertainties by investigating the biological and operational drivers of CF densification and advancing technologies that can promote and sustain it in full-scale systems.

Integrating data-driven models with domain knowledge

Current wastewater treatment plants often operate with high safety margins and excessive resource consumption, mainly due to the reliance on manual control strategies and the underutilization of available process data. Our research seeks to address this gap by developing novel modelling frameworks that combine data-driven models (DDMs) with domain knowledge (DK). Such integration can improve the model’s performance, diagnostic and explainability power, and applicability for real-world applications. This modelling approach can be adopted for soft sensing applications, development of diagnostic tools and process-relevant decision-support systems, DK-informed online control strategies that move beyond black-box predictions toward interpretable, plant-relevant guidance.

Mitigation and Quantification of GHG emissions from wastewater treatment plants

There is a high degree of uncertainty in the current methods used to quantify greenhouse gas (GHG) emissions from wastewater treatment plants (WWTPs). This uncertainty stems from non-standardized measurement protocols, limited attention to the dynamic nature of treatment processes, and a lack of tools capable of capturing these complexities. Our research lab is working to address these gaps, with machine learning (ML) playing a central role. We aim to develop improved methods for quantifying emissions by integrating dynamic process data, enabling more accurate modelling and diagnostics. This enhanced quantification is a necessary step toward mitigation, whether through smarter operational strategies or the design of bioprocesses that can reduce or even recover these emissions. GHG mitigation, both in theory and practice, remains a core area of focus for our group.