Sujit N
Hello there! I am senior machine learning engineer who loves to build end to end products integrated with machine learning tech. I have worked with multinational companies as well as cutting edge startup companies across globe. I love to take complete ownership in product development lifecycle. I am very well versed with tech stack includes C++, Python, Docker, Pytorch, TensorRT, CUDA, Nvidia DeepStream, NLP (RNN, LSTM, transformers, etc), computer vision (CNNs, Faster RCNN, Yolo model series). I have build and scaled computer vision applications (using object detection model such as Yolov5, YoloX) across more than 1000+ cameras.
REVIEWS FROM CLIENTS
I work as senior machine learning engineer at Adagrad AI. Responsible for end to end model training, deployment and integration in pro...
I work as senior machine learning engineer at Adagrad AI. Responsible for end to end model training, deployment and integration in product. Following are few of projects I have worked on:
- Person ReID & Matching- Component involves multicamera tracking. Feature vectors are generated from cropped images from multiple cameras. Feature vectors are matched to generate unique identification for each track. Custom processing involves generating metrics such as dwell-time per zone etc. Tech stack: Python, DBSCAN clustering algorithm, Docker.
- DNStar - A high performance Compute Vision inference component written completely in C++. Its responsible for performing inference in FP16 mode. Integrated object detection algorithms are implemented such as YoloV4, YoloX are integrated for performing inference and tracking using DeepSort. Pipeline supports inference in both realtime and batch processing mode. Tech Stack: Python, C++, Nvidia DeepStream, TensorRT, Docker
- GateGuard- A vehicle analytics based boom barrier automation system to automate process of vehicle authorization and barrier management using computer vision based solution. Vehicle number plate detection and character recognition is performed using YoloX object detection algorithm. Additionally contributed in development of custom line/polygon based analytics component for performing line crossing, in/out of polygon etc. analytics for custom business logic integration. Tech Stack: Python, C++, Nvidia DeepStream, TensorRT, Docker
- TaxBot - Chatbot developed using RASA framework for performing question answering related to tax related queries. Integrated Sentence BERT model for performing similarity search for out of context questions. Tech Stack: Python, RASA, Sentence BERT, Transformers
Involved in end to end development of software product. Worked in team size of 4 members. Involved in software development process fro...
Involved in end to end development of software product. Worked in team size of 4 members. Involved in software development process from requirement gathering, development and deployment of applications.
Contributed in development of Inference engine component for analysing insurance documents to parse table contents, entities and perfo...
Contributed in development of Inference engine component for analysing insurance documents to parse table contents, entities and perform correlation between them. Trained BiLSTM model in Pytorch and got F1 score of 0.91. Integrated trained pytorch model as custom component in Inference pipeline developed in Spacy.