Beyond the Model: Scaling Applied AI for Water and Resource Recovery | Nina Kshetry | ADIA Lab 2026

Value of the Information & Strength of the Argument

The talk provides valuable insights from real-world deployments, offering concrete examples of successful AI implementations in water treatment. The argumentation is compelling, built on a clear narrative that moves from generic challenges to specific solutions, using case studies to illustrate key points. The speaker’s emphasis on data quality, contextualization, and safety mechanisms (like the one-class classifier) is well-argued and practically grounded. However, the presentation would benefit from more quantitative evidence, such as specific performance metrics or comparative baseline data, to strengthen its persuasive power.

Scientific Rigor, Source Quality, Title Accuracy

The scientific rigor is moderate; while the speaker references reputable sources like the World Resources Institute and UN Sustainable Development Goals, she does not cite specific peer-reviewed studies or provide external validation for her claims. The title accurately reflects the content, and the structure of the talk is coherent, but the lack of formal citations reduces its academic robustness. The talk is more of an experience-based expert report than a systematic review or original study.

The title accurately reflects the content, emphasizing the transition from building models to scaling applied AI in water infrastructure, with a focus on implementation challenges.

7/10

Though based on extensive practical experience, the talk lacks explicit citations to peer-reviewed research or independent validation. The speaker's credibility is strong, but the evidence is anecdotal, limiting the reliability score.

Markers derived by PSI from the transcript: the creator did not define chapters.

• 00:30Introduction and overview of wastewater treatment industry scale in the US and globally.
• 03:00Discussion of water stress and key drivers for AI adoption in water facilities.
• 07:00Explanation of data sources in water facilities: smart meters, sensors, digital twins, and operator logs.
• 11:00Current state of digitization and AI adoption; most facilities still at dashboard and BI level.
• 15:00Case study in Oman: digitizing data via mobile web forms without SCADA, enabling benchmarking and KPI tracking.
• 20:00Case study in Chicago: predicting odor events with machine learning to guide mitigation.
• 25:00Case study for US Army: classifying wastewater safe to reuse using one-class classifier and safety bias.
• 30:00Real-time implementation in Pune, India: MBR health monitoring, anomaly detection, and 30% energy reduction.
• 35:00Renewable natural gas project with digital twin for process optimization and expansion planning.
• 40:00Conclusion: importance of data ontology, user trust, and AI for operator training.

• ADIA Lab Symposium 2026 — Official symposium page providing context for the event and speaker.
• World Resources Institute - Water Stress Map — Mentioned as the source for water stress data used to illustrate global water scarcity.
• UN Sustainable Development Goals — Referenced in relation to sanitation and water access goals affecting infrastructure needs.

• World Resources Institute — Corroborates the global water stress context presented in the talk.
• UN Sustainable Development Goals — Supports the framing of water infrastructure challenges in alignment with global development targets.