Machine Learning Computer Vision

Ronak Haresh Chhatbar

Æ 716-***-**** [*************@*****.*** ronak-haresh-chhatbar github.com/alphapibeta alphapibeta.github.io Professional Summary

Experienced in GPU performance optimization and machine learning technologies with a focus on profiling and improving AI models’ efficiency. Proficient in Python, C/C++, and GPU-accelerated computing, with a strong background in AI concepts including deep learning and computer vision. Skilled in multi-node configuration analysis and performance benchmarking, con- sistently enhancing system throughput and reducing bottlenecks. Education

• University at Buffalo, The State University of New York Buffalo, NY Masters in Computer Science; GPA: 3.4/4.0 Aug 2022 - Jan 2024 Courses: Operating Systems, Analysis Of Algorithms, Biometrics Image Analysis, Reinforcement Learning, Computer Vision.

• Jawaharlal Nehru Technological University Hyderabad Hyderabad, India Bachelor of Computer Science; GPA: 3.6/4.0 Aug 2015 - May 2019 Courses:Machine Learning, Cloud Computing, DSA, Computer Networks, Probability, Statistics, Mathematics, Compiler Design. Skills Summary

• Programming & Development: Python, C/C++, Java, Rust, SQL, Scala; Flask, Django; REST, GRPC APIs; Microservices; Agile Development

• GPU & AI Technologies: Deep Learning (PyTorch, TensorFlow), CUDA, GPU Performance Optimization, ONNX, TensorRT

• Cloud & DevOps: Amazon Web Services (AWS), Docker, Kubernetes, Continuous Integration (CI)

• Performance Benchmarking: Workload Profiling, Performance Monitoring, Data Loop Automation, AI Model Scalability Analysis Experience

• Tensorgo Technologies Hyderabad, India

Computer Vision Engineer Sept 2020 – Aug 2022

– Led the development of compliance software, integrating ASR technology to accurately segment and identify speakers, enhancing real-time decision metrics which resulted in a 16% increase in meeting analytics accuracy and client decision-making

– Strategically integrated eye-gaze and emotion deep learning models with Nvidia-TensorRT and Deepstream, aligning with tensorrt backend optimizations, resulting in a 40% inference boost and 25% greater system throughput for compliance software

– Refined a heart rate estimation system to address demographic diversity, utilizing BP4D+, UBFC-1, and UBFC-2 datasets, which resulted in an 8% enhancement in accuracy for a more inclusive and reliable application compliance Software

– Enhanced Agile sprints by automating model training with a multi-container Docker setup, enabling consistent, end-of-sprint deliverables that accelerated integration and performance evaluation for diverse ML, CV and ASR applications

• Wavelabs Technologies Hyderabad, India

Machine Learning Engineer May 2019 - Aug 2020

– Developed an AI-based weapon detection system for doorbell cameras using Jetson Nano, enabling on-device processing and instant threat notification to user mobile apps, achieving under 2-second identification with 30-40 FPS model performance

– In a dynamic Agile Scrum environment, managed bi-weekly sprints for a weapon detection project, leading data collection, augmentation, and model training processes, which culminated in regular sprint-end enhancements to the AI’s threat detection performance

– Implemented a real-time sentiment analysis system using ULMfit language modeling and Flask, analyzing over 5,000 customer interactions monthly and enhancing customer service quality by 6% with accurate satisfaction scoring

– Streamlined AI model training for weapon detection with Docker and AWS Sage Maker, enhancing resource efficiency by 35% and deployment speed by 20%, in sync with Agile sprint cadence for consistent sprint-end outcomes

• Wavelabs Technologies Hyderabad, India

Computer Vision Research Intern Nov 2018 - Apr 2019

– Advanced object detection accuracy by 16% through extensive experimentation with diverse architectures, optimizers, and custom neural-nets in TensorFlow, underpinned by a dataset of 3000+ manually labeled images

– Designed and implemented a facial recognition system with 95% accuracy, utilizing ResNet50 and HOG for feature extraction, and deployed age and gender classification models for real-time analytics on an i5 processor, achieving 8-10 FPS Academic Experience

• Spatial AI & Robotics Lab Buffalo, NY

Graduate Research Assistant Dr.Chen Wang — May 2023 - Present

– Achieved dual enhancements in AI vision systems by converting visual odometry models to ONNX and refining optical flow estimation, leveraging C++ plugins and mixed precision for a 33% efficiency gain and seamless TensorRT integration on Nvidia platforms

– Led backend development for robotranking.com, a platform for robotics research assessment, showcasing leadership in project management and technical innovation within the robotics community.

– Conducted detailed profiling of AI models for visual odometry, achieving significant improvement in inference speed and efficiency.

– Collaborated on multi-node GPU setups for large-scale AI workloads, enhancing overall system performance and reliability.

– Implemented GPU-accelerated computing techniques, leveraging mixed precision and CUDA optimization for advanced AI applications.

Contact this candidate