Aditya Surendran

Problem

Can gene expression data alone provide enough signal to distinguish early vs late stage before full imaging/pathology completion?

What I Built

A constrained ML pipeline trained on public RNA-seq datasets with explicit exclusion of staging inputs to isolate pure molecular signal.

Methods / Stack

Python, Pandas, ML classification workflow, independent cohort validation, low-power inference demonstration.

Key Contribution

Not “max accuracy,” but quantifying the biological ceiling of gene-expression-only staging and compressing usable signal into smaller practical panels.

Problem

Traditional BFT designs can be communication-heavy and brittle under changing network behaviour.

What I Built

An adaptive BFT framework with signature aggregation and mode-aware operation (fast/slow paths) to preserve safety while reducing overhead.

Methods / Stack

Byzantine consensus simulation, BLS aggregation concepts, adversarial jitter testing, performance instrumentation.

Key Contribution

Exploring practical communication reduction and deterministic adaptation under stress conditions rather than static protocol tuning only.

Problem

Transport optimisation often ignores health and environmental externalities in objective design.

What I’m Building

A multi-objective optimisation framing for transport decisions balancing efficiency with cleaner-climate and public health constraints.