Research Paper Implementations

Implemented and experimented with core ML/DL architectures from research papers across vision, language, and reinforcement learning.

Classical Machine Learning Projects

Built complete ML pipelines covering data exploration, preprocessing, modeling, hyperparameter tuning, and evaluation.

Micrograd from Scratch

Implemented a minimal automatic differentiation engine to understand backpropagation deeply.

Neural Networks from Scratch (NumPy)

Built a neural network framework using NumPy with layers, activations, and optimizers.

Spanish EIT Transcription Pipeline

An automated pipeline that transcribes Spanish Elicited Imitation Task (EIT) audio recordings using Faster Whisper and Silero VAD, then matches each transcription to a set of 30 target stimulus sentences using Levenshtein distance — outputting structured results to Excel for linguistic analysis.

Building a machine-learning taxon classifier for genomic classification in malaria mosquitoes

The pipeline combines MAF-based filtering with Sparse PCA (Truncated SVD) to efficiently reduce high-dimensional genomic data while preserving population-level structure. The resulting representations are then used by a 1D CNN to learn structured patterns and perform robust multiclass species classification.