Developer working at the intersection of product engineering and applied ML, from handwritten Hindu-Arabic math OCR to large-scale alignment of multilingual documents.
Stack
Selected work
Classifying emotional states from speech by fusing transfer-learned spectrogram features from Inception-Resnet-V2 with classical MFCC and LPCC coefficients.
A web and mobile app that reads handwritten Hindu-Arabic mathematical equations from a photo, recognizes their structure, and solves them.
Named-entity recognition and NLP tooling for extracting structured information from text.
Production web apps across modern frameworks, focused on performance and maintainability.
Sentence-level alignment across large parallel corpora of multilingual documents, turning unstructured bilingual archives into clean, paired training data.
About
I move between product engineering and applied research, comfortable in a Next.js codebase, a PyTorch notebook, or a messy data pipeline that needs to scale.
My past work spans Hindu-Arabic handwritten mathematical OCR, speech emotion recognition, named-entity recognition, and sentence-level alignment of bulk multilingual documents. The thread tying it together: turning ambiguous, real-world signal into something a system can act on.
Currently looking for problems where careful engineering and applied ML meet.