Rohsha T.

Mathematical AI Architect | PyTorch & Custom ML Solutions Many developers can import a library. Few can tell you why the gradient is exploding in your latent space, or how to manually optimize a loss function for custom data constraints. I am a Mathematical AI Architect specializing in building robust, high-performance Machine Learning and Deep Learning pipelines from first principles. With an MPhil in Mathematics (Double Gold Medalist) and 7 years of experience teaching foundational engineering math (Linear Algebra, Calculus, Numerical Computing) to Computer Science tracks, I bridge the gap between complex mathematical theory and production-ready PyTorch code. If your project requires bespoke neural architectures, strict data normalization, or custom algorithmic logic instead of cookie-cutter API wrappers, I design and implement them from scratch. My Engineering Arsenal & Core Capabilities: Deep Learning & Computer Vision: Custom 2D CNNs, Multi-Layer Perceptrons (MLPs), feature extraction, and latent space embeddings using PyTorch. Machine Learning Pipelines: Vectorized matrix operations, custom Gradient Descent engines, Random Forests, and hyperparameter optimization (GridSearchCV). Unsupervised Learning: High-dimensional behavioral clustering (K-Means) and dimensionality reduction (PCA). Data Engineering: Structural cleaning, feature scaling, and statistical data imputation via Pandas and NumPy. Sample Frameworks & Tools: Python | PyTorch | Scikit-Learn | NumPy | Pandas | MATLAB | Git Whether you need to architect a custom forensic engine, optimize a sequence model, or translate a complex mathematical paper into clean, modular Python code, let's connect to build a mathematically sound solution for your business.