Ankit Berwal

Designed and optimized CNN models (LeNet, VGG16) for handwritten Hindi (Devanagari) recognition using 100K+ samples, achieving ~90% accuracy. Enhanced results by 20% via CLAHE, adaptive thresholding, and segmentation. Built a real-time path-following prototype with Python, OpenCV, and Raspberry Pi using HSV filtering, Canny edges, Hough Transform, and PID control, achieving 98% path stability at 2 m/s.

Designed interior/exterior door handles from A-surfaces using CATIA V5 & SolidWorks with GD&T and tolerance stack-up. Led DFMEA & DFM/DFA reviews, cutting prototype cycles by 30%. Supported validation for thermal, vibration & endurance tests, ensuring 100% OEM compliance. Prepared technical docs & ECRs for suppliers/customers. Optimized processes for injection-molded & die-cast parts, improving efficiency and reducing cost.

Led dimensional quality checks improving tool life by 25%. Collaborated with 6 engineers to design and validate latch assemblies in SolidWorks, achieving 1.5× load capacity via ANSYS FEA. Conducted collision and packaging studies in CATIA DMU for seamless integration. Standardized inspection templates and applied SPC for tolerance monitoring and continuous improvement.