Teyj Krishnan

CNN-based computer vision pipeline for simulating low-radiation chest X-ray images. Applied denoising models implemented in PyTorch to reconstruct medical images while maintaining diagnostic quality, validated on the MedMNIST dataset with strong perceptual metrics.

Client-facing mobile application built with React Native delivering real-time mortgage market insights. Implemented features for comparing mortgage plans and calculating prices with responsive design using Figma, streamlining the decision-making process for prospective homebuyers.

EEG signal classification system for motor restoration research using PyTorch and NumPy. Processed over 350,000 EEG samples to achieve 98.15% accuracy in predicting motor intentions. Published in Journal of High School Science and awarded 1st Place at Washington State Science Fair.

High-throughput feature flag service built with Python and FastAPI, handling over 700 requests per second with sub-12ms latency. Implements a three-tier caching architecture with Redis, PostgreSQL, and in-memory fallback for graceful degradation. Uses MurmurHash3 consistent hashing to ensure deterministic user bucketing across feature rollouts.

Quantum walk simulator using Qiskit and NumPy to model network diffusion dynamics. Implements a modular Python pipeline with NetworkX and visualization tools to compare quantum and classical diffusion behaviors through entropy, coverage, and distribution metrics.

Deep learning system for classifying spinal degeneration severity from MRI scans. Trained CNN, EfficientNetV2, and ConvNext models on 147,000+ scans to achieve high AUC across multiple degeneration levels. Deployed as a full-stack FastAPI web application for real-time clinical inference.