United States

AI/ML Engineering

Master's in Electrical Engineering

Vashishtha Patil

Senior Machine Learning Engineer

Problem-solving

Critical-thinking

Empathy

Leadership

Pandas

Python

PyTorch

Java

Jupyter Notebook

Keras

Numpy

TensorFlow

Designing biometric machine learning algorithms for Amazon Halo healthtech devices

- Leading Biometrics ML algorithm development and integration with next generation of Amazon's wearable devices.

- Computer Vision frameworks and acceleration for Amazon's Smart Home Security camera devices.
- Building on-device Body composition CV/ML algorithms for Amazon Halo to accurately measure body fat percentage using smartphone camera.
- Building Tone Analysis feature for Amazon Halo wearable device that helps customers measure, understand, and improve their health every day.

• As part of algorithm development team in Qualcomm's Computational Photography group, I am responsible for prototyping, developing, optimizing and commercializing new single and dual camera computational imaging features to drive mobile camera features beyond the traditional ISP.

• Designed Deep Learning ARM CPU inference framework for image enhancement features such as super-resolution, portrait segmentation, etc.

• Developed and optimized dual camera image fusion algorithm, “Qualcomm Clear Sight[1]” which blends images from Bayer & Monochrome sensor to improve picture quality (dynamic range and sharpness with less noise in low light) while meeting stringent latency & accuracy requirements

• End-to-end design and implementation of Stereo based dense depth map computation
• Developed & optimized depth effects such as realistic bokeh (portrait mode), post-capture refocus & segmentation [2]

• Implemented fixed point design using NEON vectorization for several image processing algorithms such as edge preserving filters (Guided Filter, Domain Transform Filter), Weighted TVL1, Alpha blending, Joint up-sampling, LoG, etc.

TECHNICAL SKILLS
• C++, C++11 features & principles, OpenCV, ARM NEON vectorization, parallel programming,
Camera ISP, GIT, Android NDK, debugging/performance analysis of native Android libraries, clean and effective coding, being able to look at the big picture, processing information quickly and getting to the gist, working with remote teams, etc.

[1]: https://goo.gl/9vuXmu 
[2]: https://goo.gl/nbfCef

• Developed motion compensated multi-image temporal denoising algorithm and low-light camera features

• Implemented various enhancements to reduce shot-to-shot camera latency and improve customer experience of several camera features

• Worked on product integration on Snapdragon chipsets with several top smartphone and connected camera OEMs

• Worked with Qualcomm legal teams to release our features outside Qualcomm

• Developed demo Android apps using Camera2 API to showcase Computational Photography features to OEM customers at Qualcomm internal events

[1]: https://goo.gl/CsZuLh

• Designing, developing and implementing test plans, cases and test processes for integration and testing of various image processing and computer vision technologies
• Preparing test data for thorough testing and analysis of the test cases/ plans
• Implementing new metrics to aid Image Quality Assessment for computational photography projects
• Identifying and documenting computer vision software bugs and procedures to enable it’s accurate replication
• Developing android testing tools for automated regression testing, error/bug retests and usability of computational photography’s android applications

Development of image analysis algorithms and software to analyze droplet dynamics during its generation process.

Development of software to analyze diagnostic videos and extract quantitative measurements.

Project: Tracking Tissue Movements of Lungs due to Breathing Pattern
* Developed an open loop respiratory tracking system (RTS) to track tissue movement of Lungs by analysing breathing
pattern before and during critical procedures such as biopsy
* Programmed and implemented the system on DSP platform TMS320C6713 using Code Composer Studio

Vashishtha Patil🇺🇸

About Vashishtha Patil

Expertise

Sessions with Vashishtha Patil