Krupal Purohit Email & Phone Number

Madrid, Es

Gandhinagar, Gujarat, India

I am currently working on pre-training and fine-tuning LLM like LLama, Flacon, etc. on custom data for clients. Deployed LLM on client's infrastructure for efficient inference and data privacy. Used quantization techniques to reduce the memory footprint of the model.Worked on Document Q&A service(cognitive search), which updated existing keyword-based search pipeline to embedding-based search pipeline using transformer models and elasticsearch. Led a Team of 4 developers, developed… Show more I am currently working on pre-training and fine-tuning LLM like LLama, Flacon, etc. on custom data for clients. Deployed LLM on client's infrastructure for efficient inference and data privacy. Used quantization techniques to reduce the memory footprint of the model.Worked on Document Q&A service(cognitive search), which updated existing keyword-based search pipeline to embedding-based search pipeline using transformer models and elasticsearch. Led a Team of 4 developers, developed a document Q&A system for internal company use from scratch, and was also responsible for deploying it using Kubernetes. Working on audio and text analysis systems, the primary goal of the system was to help HR filter out candidates with good communication skills. Developed it, integrated it, and deployed it using Kubernetes. Led a team of 2 and developed JD-CV comparison, Deployed a few modules for the JD-CV comparison project. Show less

Gujarat, India

As an AI ML Engineer, as my first project, I developed and deployed a chatbot for the client from scratch using Bot Framework and Azure bot service. Learned a lot about project management, client communication, software architecture, software development, and deployment. As the next project, I worked on a POC for extracting table data using OCR and OpenCV. As Team Lead, developed a second chatbot for the client and learned about team leading, mentoring, and task management. Next… Show more As an AI ML Engineer, as my first project, I developed and deployed a chatbot for the client from scratch using Bot Framework and Azure bot service. Learned a lot about project management, client communication, software architecture, software development, and deployment. As the next project, I worked on a POC for extracting table data using OCR and OpenCV. As Team Lead, developed a second chatbot for the client and learned about team leading, mentoring, and task management. Next worked on a resume parser, in which trained a document layout model along with an NER model to extract relevant data in a structured format.Next worked on a rasa-based chatbot for clients, in which trained an intent and entity detection model, also did the re-architecture of the entire chatbot to support dynamic chat flow creation. Successfully completed and deployed it. Show less

Ahmadabad, Gujarat

Joined as a Lead Python AI Developer and took Ownership of a computer vision-based project. Worked on replacing the old opencv-based video processing pipeline with a new deepstream-based pipeline. Got experience working with Nvidia Nano and Nvidia TX2. Connected with the client, identified the system's key goal, and developed new modules to achieve those. Collected data, cleaned, and preprocessed for training. Trained different models such as object detection, object… Show more Joined as a Lead Python AI Developer and took Ownership of a computer vision-based project. Worked on replacing the old opencv-based video processing pipeline with a new deepstream-based pipeline. Got experience working with Nvidia Nano and Nvidia TX2. Connected with the client, identified the system's key goal, and developed new modules to achieve those. Collected data, cleaned, and preprocessed for training. Trained different models such as object detection, object classification, and object tracking. Got experience with messing queues such as NATS, and rabbitmq and learned about different types of video streaming. Show less

India

Joined as an intern, during which worked on writing scripts to clean and process audio data for training. After 4 months of the internship, continued working on different models such as speech-to-text, text-to-speech, Text Classification, and Sentiment analysis. Developed custom modules for streaming inference and training. Got ownership for the End-To-End speech-to-text process, which successfully did R&D and moved from Kaldi speech-to-text to deepspeech speech-to-text.