Akash Desai

Working on Object detections & object tracking, Face, Age, Gender, Race detection Models. Improving existing model accuracy. Model optimization, video analytics inferencing, optimisation of model - Onnx, tensorRT. Working on various object detection models like yolov4, yolor, yolo-csp, mish, efficienetdet, efficientnetv2, swin transform. etc for better accuracy. working on model training to testing the pipeline. model training on AWS server, linux & window system setup. Research, design, implementation & evaluation of novel computer vision models Working on large-scale datasets & focusing on creating scalable & accurate computer vision systems.

Objective was to Detect and Monitor various warehouse Equipment for optimal working. Utilized web scraping techniques to extract and organize client data. Reviewed image quality and annotations. Worked on Object detection & segmentation using state-of-the-art models (Tensorflow1, 2, YOLO, Detectron2). Worked on Model conversion (Tensorflowlite, Onxx etc).

Developed an NLP solution to identify types of customer complaint & assigning team to solve the problem. Used Text mining using NLP to segregate the type of queries and complaints received. Trained on 5 years of data, achieving 89.2% accuracy. Used NLTK & Spacy libraries for extracting useful information. Worked on NLP for Sentiment Analysis. Worked on Machine learning (RF, Boosting, Ensemble, LSTM, BERT models) to increase model accuracy. Worked on Language Modelling (RNN, ELMO, GPT2) for automatic text completion.