RunAs Radio Sustainable AI with Darshna Shah
10 snips
Mar 18, 2026 Darshna Shah, Chief AI Officer and former neuroscientist who builds ML at scale, talks sustainable AI. She covers how big model experiments strain power grids. She explores designing for efficiency, measuring consumption and lifecycle emissions. The conversation looks at evaluations, safety in retrieval systems, and innovations to shrink data center impact.
AI Snips
Chapters
Transcript
Episode notes
Lab Work With Excised Brain Tissue
- Darshna Shah described taking excised human brain tissue from surgery to her lab, implanting electrodes, and running experiments to study drug-resistant epilepsy.
- She used whole-cell patching and network recordings to probe excitability and infer why drugs failed for children with 10–20 seizures daily.
Bootcamp That Converted A Neuroscientist To Data Science
- Darshna recounted a data science bootcamp that transitioned her from academia to industry by giving real-world projects and a fast-paced finance engagement.
- Her team used unsupervised learning to analyze asset manager behavior and produced actionable results within weeks, which hooked her on applied ML.
Show LLM Confidence To Change Trust
- Darshna noted that LLMs hide deterministic confidence and urged surfacing confidence scores to change user trust.
- Showing confidence alters user interpretation and willingness to trust generative outputs in UI design.