Shreyas Rajesh Email & Phone Number

Santa Monica, California, United States

San Diego, California, United States

Served as teaching assistant for- ECE 5: Introduction to Electrical and Computer Engineering (FA20, WI21)- ECE 17: Object Oriented Programming: Design and Development with C++ (SP20, SP21)Responsibilities include designing programming assignments and exams, machine learning lectures, follow along jupyter notebooks and weekly discussion/problem solving sessions.

San Francisco Bay Area

Assisted across the stack of the Machine Learning Pipeline for Computer Vision Problems with significant developments on both the applied research as well as the engineering fronts. - Designed and built an end to end Machine Learning Pipeline for Computer Vision problems from scratch as part of a small engineering team- Designed a frontend dashboard using Material UI, React and Redux for machine learning model building (metrics tracking) and deployment on images and videos with AWS components.- Introduced a human pose estimation service into the machine learning pipeline for model setup and deployment.- Implemented an object tracking service using deep learning models for auto-annotation using a web-based annotation tool handling both the frontend as well as backend implementations for the same while taking sole ownership for the applied research and engineering efforts.- Incorporated and maintained very high code quality and software engineering practices during this work.

Bengaluru Area, India

Worked as a pre-dominantly R&D engineer for the company for their new projects. DL Frameworks: PyTorch, Darknet, TensorFlow, KerasLibraries and languages: Numpy, Scikit, OpenCV, Python, C++Version Control and Project Management: Git, Bitbucket, Visual Studio Team Services(VSTS)Represented the company at the Qualcomm Smart Cities Accelerate conference 2019 at San Diego manning the booth and interacting with investors, VCs and CXOs from across the Globe.POSE ESTIMATION:- Worked on developing high accuracy pose estimation model for realtime inference on the edge.- Research on state of the art pose estimation methods and techniques and development of new model that performs on the same level in terms of accuracy but provides accelerated runtime using the best and quickest features of various state of the art models simultaneously. - Developed a new custom pose algorithm for our client's surveillance use-case.- Retrained, modified and optimised existing state of the art pose estimation models to deploy on the edge on devices like NVIDIA TX2, Intel Upsquared etc.- Optimised custom pose estimation models using optimisation libraries for deployment on respective edge devices.- Integrated pose estimation module with various other deep learning modules and helped develop end to end pipeline with ZMQ communication and dockers.MISC:- Helped with pipeline setup and deployment of product on cloud for ATM surveillance tasks for over 300 cameras. - ZMQ and custom architecture integration as well as modularisation of code using docker. - System and environment setup using docker for various tasks.- Trained Classifiers and Object detectors for various tasks using CNNs and worked on R&D for improvement of the accuracy of models especially in special cases like class imbalance etc.

Bengaluru Area, India

Internship Program for working professionals towards improvement of Machine Learning skills including two capstone projects that involved full implementations of research papers and improvement on these research results.- Realtime 2D face reconstruction: Re-implemented 2D face reconstruction in Tensorflow from CVPR 2018 paper, "SfSNet: Learning Shape, Reflectance and Illuminance of Faces ‘in the wild’" by Sengupta et al, originally in Caffe with incremental performance.-Automatic Photo Styling using Style Transfer: Reimplemented ”Fast End-to-End trainable guided filter” a paper from CVPR 2018 in keras with a jupyter notebook version for the same.

Singapore

Computational Fluid Dynamics of embryonic/fetal cardiac flow at the Cardiovascular Biomechanics and Ultrasound laboratory.- Published a paper titled Fluid dynamics and forces in the HH25 avian embryonic outflow tract as a co-author- Worked with the Vascular Modelling Toolkit to model the vasculature of the avian heart from ultrasound scans. - Gained experience with MATLAB and C for geometric modelling and curve fitting of the various parts of the avian heart. - Worked on running simulations of the outflow tract in the heart to observe and study the flow patterns of blood in the heart of the avian outflow tract using ANSYS fluent and C programming. - Exposure to the research process, reproducing results for multiple cases, analysing results and making decisions on future research directions as well as overcoming roadblocks mainly due to lack of sufficient data and resources for multiple test cases. - Experience in the research article writing and review process for an international peer reviewed journal.

Teacher's assistant(including lab work) for the course Engineering Graphics based entirely in AUTOCAD.

Mysuru Area, India

Towards Practice School - 1 program of BITS PilaniAutomation of rubber moulding process : - Involved innovative and new model design and assembly using SolidWorks for rubber moulding process- Understanding of the working of a mechanical industry and experience on shop floor- Understanding of the problems and drawbacks with excessive dependence on manual labour and necessity for automation in the industry.

Master Of Science - Ms, Machine Learning And Data Science

Bachelor Of Engineering - Be, Mechanical Engineering, 8.32/10

Activities and Societies: Basketball - College Team Vice Captain Aerodynamics ClubEarly Graduate - Completing degree with 3 years.