Vedang Kshirsagar

Architected and shipped an end-to-end 3DMM face reconstruction system using PCA-driven models, cutting 3D reconstruction error by 30% through architecture iteration on CNN/Transformer hybrids. Cut model inference latency by 40% via quantization, ONNX conversion, and runtime optimizations, enabling real-time avatar processing at scale.

Built the Windows desktop client (Python, Tkinter) for an arterial pulse analyzer medical device, streaming live hardware waveform data over a serial COM port into threaded acquisition loops that sliced the signal into rolling 2-second windows and rendered real-time Matplotlib plots with signal-quality feedback. Implemented the signal processing and reporting layer: Butterworth filtering and SciPy

Pursued a Bachelor of Technology in Computer Science & Business Systems, covering Python, cloud computing, database management, object-oriented programming, and data structures.