Kannan

Applied physics-informed neural networks (PINNs) to model carbon nanoclusters. Boosted prediction accuracy by 80% and cut simulation time by 60%. Designed an end-to-end GPU-accelerated pipeline in TensorFlow + CUDA.

Architected and deployed multiple iOS applications including '1minute DOEShelp', 'iPong', and 'DabCounter'. Integrated CoreML for real-time inference, improving speed by 70% on Apple Watch hardware. Managed full product lifecycle.

Developed CNNs for biomedical imaging achieving 40% improvement over baselines. Implemented RL-based trading bots using LSTMs with mixed-precision GPU training.

Built LSTM + Word2Vec NLP system to auto-assign ICD-9 codes from clinical notes, raising accuracy from 42% to 71%. Developed anomaly detection tools.