Srinivas Harish | ML, CV & GenAI

About Me

I’m Srinivas Harish — a computer engineer deeply immersed in machine learning, computer vision, and generative AI. I build systems that see, think, and adapt — from YOLO-powered real-time game analysis to MATLAB-based orbital transfer optimizations. Currently, my research focuses on improving the adversarial robustness and accuracy of LLM-generated text detection, pushing the boundaries of generative AI interpretability and safety. At NYU, I also tutor Linear Algebra and CS, bridging foundational theory with real-world impact. My mission is to engineer what endures, and innovate what matters. 🚀

Projects

⚽ Real-Time Football Game Analysis

YOLOv8 + Kalman filter + Optical Flow → Real-time vision system for football dynamics.

Fine-tuned YOLOv8 on custom datasets for player and ball detection.
Segmented players using Gaussian Mixture Models and extracted t-shirt colors.
Applied Kalman filter for robust object tracking under occlusions.
Analyzed scene dynamics via optical flow and perspective transforms.
View Project

🔋 N-Channel MOSFET Performance Research

Analyzed I_ds-V curves to understand saturation, transconductance, and nonlinearity.

Identified 3.6V threshold; modeled logistic curve behavior with saturation points at 2.0V & 10.0V.
Measured >10 Siemens transconductance before sharp drop-off post-saturation.
Research PDF

🧪 Advanced Acetaminophen Synthesis Optimization

Refined 24-step synthesis pipeline; achieved 93.8% yield with 170.2°C purity.

Optimized molar ratios and removed unnecessary filtration/acidification steps.
Proven significance (p < 0.00001) over 16 experimental trials.
Research PDF

🚀 Hohmann Transfer Trajectory Optimizer

Simulated 30M+ launches to find optimal Earth-Mars transfer windows.

Used MATLAB and ODE solvers to simulate delta-v across 36,500 days.
Refined ideal launch to Oct 2, 2024 with Δv of 5816.1 m/s.
Research PDF

🔌 Full-Bridge Rectifier Optimization

Reduced ripple voltage by 96% via optimized capacitor choice in diode circuits.

Analyzed ripple reduction across 0–100μF, 10–50V with Schottky diodes.
Achieved R² > 0.99 using X5R MLCC capacitors.
Research PDF