Ram Dyuthi Sristi

About Me

I am a final‑year Ph.D. candidate in Electrical Engineering at UC San Diego, advised by Prof. Gal Mishne, designing interpretable machine‑learning frameworks that weave multimodal biomedical signals into detailed maps of brain connectivity. My work has appeared in NeurIPS and ICML and builds on deep generative models—Transformers, VAEs, and coupled autoencoders—combined with rigorous statistics. Earlier, I graduated top of my class at IIT Hyderabad, earning the Institute Silver Medal and a Research Excellence Award for my EMBC’17 machine‑learning wearable‑sensing study. I mentor students and co‑lead brain‑computer‑interface workshops, aiming to transform algorithmic advances into real‑world biomedical impact.

Research Highlights

Conditional Stochastic Gates

Context‑aware feature selection

c‑STG learns which input features matter for each context, boosting accuracy and interpretability across domains.

Paper ↗

Complementary Cortical & Thalamic Inputs

Dissecting cell‑type‑specific striatal dynamics

Large‑scale recordings reveal how cortex and thalamus jointly shape direct‑ and indirect‑pathway MSNs during skilled locomotion.

Pre‑print ↗

Differential Spectral Clustering (DiSC)

Finding feature clusters that separate conditions

DiSC spots groups of pixels/genes/voxels whose relationships shift between states, outperforming baselines on imaging and genomics.

Paper ↗

Coupled Transformer Autoencoder

Disentangling shared vs. private neural dynamics

CTAE captures long‑range, non‑linear interactions across brain regions with explicit shared/private latent spaces.

Under review (ICLR 2026)

Teaching & Outreach

Co‑organised Brain‑Computer Interface workshops (UCSD 2023 · UMich 2024, ≈40 participants each).
Mentor in Simons Collaboration on the Global Brain SURFiN program (2022‑23, 2024‑25).
Teaching Assistant — DSE 220: Machine Learning · CSE 151A: Intro to Machine Learning.
Guest talk at FinDS (UCSD): “Seeing data clearly – the art of visualisation” for high‑schoolers.
Reviewer: NeurIPS, ICML, ICLR (2024–2025).

Interpretable ML × Neuroscience