Abhishek Tomar

Developed a robust, high-traffic question search system handling over 2 million req/day, integrating Similar Question and Auto Completion modules using ElasticSearch, Transformer(Bert), ANN, and Learning-To-Rank model. Developed universal search feature utilizing Bert Transformer for query intent recognition and entity extraction (F1 score of 0.96). Developed hyper-personalised products using Large Language Models (LLMs) like GPT4, with prompt engineering, Langchain, and Pinecone. Evaluated and fine-tuned open-source LLMs (LLaMA, LLaMA2, Falcon) using QLoRA and Peft framework. Built a cGAN-based model for watermark removal and fine-tuned YOLOv5 for upper body object detection.

Developed a call assignment system for Health Care Practitioners across 28 countries using unsupervised clustering and call grids. Developed a model to forecast sales for a new drug using regression and neural networks. Created a roster integration module for daily employee status tracking using Talend and PostgreSQL. Implemented an end-to-end trainable two-stream fusion architecture for video action classification. Developed a text classification system using Support Vector Machine (SVM), PCA, and Auto-Encoders.

Covers big data analysis using scalable machine learning techniques and introduces Apache Spark for large-scale data processing.

Provides foundational knowledge on deep learning concepts, techniques, and their application in various fields.

Explores the mathematics and application of deep learning models designed to generate new data.