Safe water scarcity affects more than a billion people daily and is the primary cause of health impacts in many off-grid communities of low and middle-income countries (LMICs). Low-C, decentralized, community-run water treatment and delivery systems are crucial for an efficient water-energy-agriculture nexus, but reliability issues and lack of regular supervision can lead to poor water recovery and O&M. My research involves real-time imaging and data-driven decision support framework design to improve that scenario.

Robust models on bulk water and salt transport across water filtration membranes exist, but mechanistic insights into scale growth in the active layers of the membrane, particularly in the incipient stages, are still lacking. Moreover, visual investigations on membrane PA-layer scaling are rare, despite their importance in advanced membrane design and operation. Hence, we developed a model framework to visualize scale formation on RO membrane active layers at mm-to-nm length scales. Mineral transport identified through the TEM–K-means protocol hinted at the broader use of this technique in designing advanced membrane materials with custom surface chemistry.

Even after addressing membrane fouling and degradation issues, the ‘unsupervised’ operation of PVRO plants, due to the lack of regular availability of trained engineers, may still result in suboptimal operation. So, I worked on designing a low-cost semiautomated supervisory control and data acquisition (SCADA) protocol for small-scale PVRO plants to improve their technology readiness level. Monitoring and forecasting PVRO performance (from historical plant data) can serve as a crucial decision-support system in such a framework. XGBoost and LSTM models were preliminarily tested, demonstrating reasonable performance for short-term predictive control of PVROs.

In-situ observation of complex fluid transport phenomena is critical for proper understanding and detailed design of engineering systems, especially in designing innovative and low-cost water treatment and sanitation systems for sustainable application in LMICs. In-situ imaging of such systems (via microscopy and macroscopic equipment) is often challenging and limited due to several constraints. Custom-built batch systems designed to facilitate those visualizations and computational post-processing by employing classical and advanced (AI and ML-based) algorithms are essential. My expertise in this area is in imaging membrane-based fouling kinetics.

Solid waste management is a critical challenge for most municipalities in South Asia, and the problems range from a lack of standardized policies to arbitrary waste collection systems as well as the failure to persuade people to separate recyclables from garbage. The COLLECT3R pilot project would simultaneously focus on (a) changing the ‘we dump – they collect’ behavior of residents and promoting source separation of residential wastes, (b) a modular source-separated (organic/inorganic) waste collection and transportation arrangement, and (c) data collection and monitoring the key performance indicators (KPIs) to evaluate the project’s success.

While commercial slurry pipelines transport a wide size range of particle mixtures, phenomenological erosion models developed by lab-scale testers are typically based on experiments with slurries having a single particle size distribution (PSD). So, pipeline erosion using single and bi-modal PSD slurries were compared to understand the effect of particle-particle and particle-wall interactions. Simulated pipeline erosion experiments were carried out using a TWT by mixing sand particles with gravel. The results demonstrated a synergistic erosion behavior; parametric studies varying the particle abrasiveness hinted that the possible formation of sand-gravel blocks and gravel-assisted sand impact may have contributed to this accelerated erosion behavior of bi-modal slurries. These new insights highlight the insufficiency of existing erosion models in interpreting the physics of wear in slurry pipelines.

In early 2020, I co-led a team of public health professionals, data analysts, and content creators to track the growth of COVID-19 in Bangladesh. Besides, my team designed an open-source python-based pipeline for sharing daily COVID-19 case updates via interactive charts, exploratory visualizations, and infographics. Besides using social media as a dissemination tool, my team also designed a free domain and android app for faster data delivery and concise (yet powerful graphical) data reporting.

Field-scale pipe erosion tests are difficult to control, so lab-scale experiments are frequently employed. However, most techniques produce hydrodynamic conditions dissimilar to pipe flow. The TWT hydrodynamics closely resembles pipe flow than other wear-testers. This work evaluated TWT’s repeatability, parametric wear trends, and the major differences between TWT and pipe-flow hydrodynamics. Visualization experiments showed that only larger particle sizes or lower TWT speeds are suitable for generating sliding-bed-dominated erosion comparable to pipe flow. In addition to previously observed particle–coupon contact time and particle degradation effects, normal load and coupon edge effects are identified as necessary. Despite these limitations, the TWT is a promising apparatus for ranking material wear resistance and, with careful interpretation, for making pipeline erosion predictions.