At AI-Shala, I designed and implemented a dependency graph-based automated grading system with explainable evaluation workflows. I also developed document understanding pipelines using OCR, document layout analysis, and structured information extraction.
At myNachiketa, I developed machine translation pipelines and style imitation models for low-resource languages, reducing manual effort by 40% and improving generation fidelity by 25%. I also fine-tuned transformer-based language models and deployed FastAPI inference services on AWS EC2.

