Dheeraj Madda Email & Phone Number

Bengaluru, Ka, In

Bengaluru, Karnataka, India

𝗥𝗲𝘀𝗲𝗮𝗿𝗰𝗵 𝗮𝗻𝗱 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗺𝗲𝗻𝘁 :-• Real-time Instance Segmentation - Performed R&D in replacing YOLOv8 Instance Segmentation model with an efficient NMS-Free model to facilitate the removal of AGPL 3.0 License for real-time application. The qualitative and quantitative results, both, were superior to YOLOv8 in segmenting each individual SKU in a given frame.• Loss function optimization - Led R&D efforts to optimize the existing loss function for classification. Improved the model's generalizability and accelerated convergence resulting in reduction of training time by more than 60% on a single GPU by just replacing the existing loss function with the proposed one.• Continual Learning - Researched and developed Domain-Incremental and Class-Incremental Continual Learning techniques, delivering substantial scalability improvements. Successfully slashed training time by over 80% for new SKU classifications, maintaining exceptional accuracy for existing SKU categories.𝗦𝘁𝗿𝗲𝗮𝗺𝗹𝗶𝗻𝗲 :-• Developed a robust, scalable PyTorch training framework tailored for diverse models, datasets, and training scenarios (e.g. Distributed Training, Knowledge Distillation, SSL, Continual Learning). Streamlined operations, reducing development cycles by more than 10x and enhancing overall training efficiency.

Hyderabad, Telangana, India

• Discussed current trends in AI with Data Science Leads and Director. Worked on implementing the same as part of Internal Computer Vision Platform.• As part of internal Computer Vision team, wrote various blogs on topics like Camera calibration, Self-supervised learning, Object tracking, Multi-task network.• Mentored and guided juniors on various computer vision projects for client use cases.

Hyderabad, Telangana, India

• Worked on a project for one of the largest Retail Companies in the US where I have built various backend APIs on Azure as per the business requirements. Secured the APIs using JWT.• Worked on implementing Databricks workloads and improved existing workloads by optimizing them for cost and memory.• As part of Tiger's Internal Computer Vision Platform, developed various modules while maintaining high coding standards and curated it to be used on different datasets, on the following topics:-a) Knowledge Distillation:Distilled large models across different tasks to optimize for less memory and low latency.b) Pruning and Quantization:Explored various pruning techniques and created a custom pipeline to Prune the CNN networks efficiently. Developed scripts for performing various Quantization techniques for CPU and GPU using frameworks - ONNX, Intel OpenVINO, Nvidia TensorRT.c) Multi-Object Tracking:Explored and compared 7 different MOT trackers for speed that are released from the year 2016 - 2023. Developed scripts that can be used for seamless integration of different MOT trackers with the object detection algorithm of choice. This led to faster development and experimentation of different trackers for client use cases.

Hyderabad, Telangana, India

• Worked with the team on Research and Development of training efficient networks in PyTorch.• Worked on optimizing deep learning models using Intel OpenVINO and improved the inference latency by 60%.• Deployed the model on AWS serverless computing platform using light-weight runtime.• Worked on implementing a conversational bot using AWS Lex and integrating it with the AWS serverless platform.

Pune, Maharashtra, India

• PoC to Production - Design, develop, deploy and secure the serverless application built entirely using AWS services that are being used by thousands of users.• Design and develop customized and scalable AWS solutions as per business requirements.• Processing large files using a serverless computing platform while building a robust solution to handle the serverless compute timeout limitations.• AWS services - API Gateway with Custom OKTA Authorizer, Lambda Functions-Python3, Step Functions, EC2, VPC, RDS, S3, SNS, SQS, EFS, CloudFormation

Post Graduate Diploma, Internet Of Things (Iot), A

Master Of Engineering - Me, Electronics And Telecommunication Engineering, A+

Bachelor Of Engineering - Be, Electronics And Telecommunication Engineering, I

12Th, Pcme, I

10Th, I