Hello, I'm
Chinmay Rane, PhD
Senior Machine Learning Engineer
drranechinmay@gmail.com
+1 682 583 1880
Austin, Texas, 78664
Senior Machine Learning Engineer
drranechinmay@gmail.com
+1 682 583 1880
Austin, Texas, 78664
Hello, I'm Chinmay, a Senior Machine Learning Engineer based in Austin, Texas, with extensive experience in developing cutting-edge AI algorithms. After earning my PhD, where I focused on adaptive activation functions across various machine learning models, I chose to work in a consultancy-based company to gain exposure to building practical AI applications. In this role, I’ve contributed to a diverse set of projects across multiple domains, including AI for rail car safety, road safety, medical imaging, vision-language models, and large language models for repair instructions. Currently, I am exploring the exciting 3D space with vision-language models.
While deeply involved in these real-world applications, I have continued to engage in research work in areas such as radiology, pathology, and drug discovery, and have published several papers in these fields. This blend of applied work and research underscores my commitment to both advancing AI technology in industry and pushing the boundaries of knowledge through academic exploration.
In addition to my technical expertise, my roles as a GTM, technical architect, and technical program manager have allowed me to hone my leadership, project management, and team collaboration skills. I have successfully managed multiple projects, delivering results on time and ensuring customer satisfaction, all while fostering effective communication between technical teams and clients.
Download Resume-Leading the development and deployment of innovative AI solutions by combining expertise in Generative AI, multimodal AI/ML across LLMs, vision-LLMs, computer vision, and medical imaging with strategic leadership and project management. Driving impactful results in both small and large-scale enterprise deployments across industries.
- Developed a custom Semantic search RAG pipeline. Can be integrated with any application with changes to model and prompts
- Developed a custom web-based RAG pipeline for Llama 3.1 and Mistral 7B, implemented optimized neural text-to-speech using Piper software and integrated Whisper for speech-to-text on Raspberry Pi using Wyoming protocol through personal server, and trained a custom wake word using open-source tools.
- Implemented the YOLOv8 object detection algorithm using Docker and Kubernetes for scalable and efficient deployment, ensuring seamless integration and optimized performance in containerized environments.
Data Scientist Intern at Unique Software Development
- My PhD from The University of Texas at Arlington was focused in developing adaptive activations for any neural networks. My research focused on the role of activation functions in neural networks, particularly in shallow CNNs, an area often overlooked compared to deeper architectures like ResNet. I developed AdAct, a novel Adaptive Activation algorithm that leverages piecewise linear properties to enhance model performance. Through extensive experiments, I demonstrated that AdAct outperforms traditional ReLU across various CNN and multilayer perceptron architectures, showcasing its potential for improving deep learning efficiency and accuracy. Dissertation.
- Worked at Image processing and Neural Network lab as a research asistant
- My MsC from The University of Texas at Arlington was focused in developing the initial version on adaptive activations for approximation and classification.
- Worked at Image processing and Neural Network lab as a research aissitant
Utilized generative AI for query understanding and response generation, developing a semantic search pipeline leveraging Milvus DB, a reranker, and a reader to enable contextual retrieval and reasoning across baby book information, human anatomy, and infectious diseases. Integrated LLM-based retrieval-augmented generation (RAG) techniques to enhance information synthesis. Optimized search relevance and accuracy by dynamically adapting prompts and models, enabling the system to deliver accurate and contextualized search across diverse domains.
Served as a Technical Program Manager to lead the development of a Vision-Language Model (VLM) for describing defective aircraft parts and a Retrieval-Augmented Generation (RAG) system powered by Large Language Models (LLMs) for generating repair instructions. Directed efforts in fine-tuning VLMs with expert-provided captions, preprocessing repair manuals through cleaning and chunking, and implementing advanced search and generation methodologies to optimize aircraft safety workflows.
Served as the Lead Technical Architect to design and implement an end-to-end pipeline for cardiac MRI, integrating AI-assisted annotation, custom self-supervised learning algorithms with multi-GPU support, and inference optimization using TensorRT and Triton Inference Server. Enhanced visualization for a user-friendly application and led the deployment of a dockerized solution on AWS using Lambda functions and EC2 instances for scalability and reliability..
This project involved optimizing a customers custom tensorflow object detection pipeline. The customer wanted to improve their inference pipeline with tensorRT and triton inference server. We use Netron, ONXX graph surgeon and optimized individual graphs based on speed and also included DALI to speed up the data loading process.
Worked on creating a custom pipeline for tensorflow federated learning using tensorflow light training so as to avoid sending data over cloud.
Working as a technical lead for an ongoing project with multiple use cases. Duties involve setting a timeline to deliver each use case, assisting in the clients annotation team, setting initial technical path, working with the team to complete the task on time and finally, involvement in the technical document delivery. Also, involved in redesigning and proposing new improved techniques for the inference pipeline by including active learning, drift detection, AI assisted annotation and self supervised learning.
This project involved two separate training algorithms, first is the prediction of DNA to RNA and second is RNA to Proteins. Traditional dimension reduction models such as SVD, PCA along with more advanced diffusion and latent diffusion models with various downstream tasks were experimented.
Authors: Kanishka Tyagi, Chinmay Rane, Ketaki Vaidya, Jeshwanth Challgundla, Soumitro Swapan Auddy, Michael Manry
ArXiv
Publish Date: Dec 2024
Authors: Chinmay Rane, Kanishka Tyagi, Adrienne Kline, Tushar Chugh, Michael Manry
Evolutionary IntelligenceThe Second Tiny Papers Track at ICLR 2024Publish Date: October 2024
Authors: Chinmay Rane, Kanishka Tyagi, Tushar Chugh, Nirmala Murali, Michael Manry
The Second Tiny Papers Track at ICLR 2024
Publish Date: March 2024
Authors: Kanishka Tyagi, Chinmay Rane, Michael Manry
ArXiv
Publish Date: August 2023
Authors: Chinmay Rane, Sanjeev Mallur, Yash Shinge, Kanishka Tyagi, Michael Manry
ArXiv
Publish Date: April 2023
Authors: Kanishka Tyagi, Chinmay Rane, Michael Manry
Elsevier
Publish Date: Late 2021
Authors: Kanishka Tyagi, Chinmay Rane, Michael Manry
Elsevier
Publish Date: Late 2021
Authors: Kanishka Tyagi, Chinmay Rane, Michael Manry
Elsevier
Publish Date: Late 2021
Authors: Kanishka Tyagi, Chinmay Rane, Bito Irie, Michael Manry
SN Computer Science
Publish Date: June 2021
Outside of work, I cherish spending time with my loving family. My wife, Stacy, a compassionate therapist, and I share a beautiful blend of Western and Indian cultural backgrounds, which enriches our lives with diverse perspectives and traditions. We have recently welcomed our newborn, who brings immense joy to our lives. Our family is also completed by our adorable dogs.
In my free time, I enjoy exploring new outdoor activities like playing soccer and volleyball, and I also have a deep passion for motorcycle riding — a love that probably started from weaving through the chaotic streets of India on a two-wheeler. I appreciate engaging in fun projects that combine my passion for AI with everyday life, such as developing AI-based semantic search tools using baby books and creating AI home security systems. Additionally, I cherish spending quality time with my family, whether it's going for walks with my wife and our dog, or now, with the added joy of our newborn, creating new memories together.
