Chat-GPT Plugins & CustomLLMs: Developed a chatGPT like bot using OpenAI LLM APIs. The chat-bot is able to retrieve real time flights/hotels data from MMT’s APIs and vector databases. Tech used: LangChain, Redis, FastAPI. (live on 100% traffic). Currently working on fine-tuning and Knowledge Distillation of open source LLMs

Leading the development and maintenance of 2 DS products. Responsible for model development, automated model training and building scalable APIs. Both products have positive PnL and show strong performance in evaluation metrics○ Flights Fare Lock: Developed a multi-objective Flights Fare Prediction model using a multi-head NN. An asynchronous API is deployed for handling the large amount of traffic. Tech used: Pytorch, MLFlow, Kafka, Aerospike, ElasticSearch○ Wait-listed ticket confirmation prediction: Developed an embeddings based ensemble NN model for predicting the confirmation probability of wait listed Indian Railways tickets. The model powers the Trip Guarantee product offered on MMT

Working on anomalous sound detection and acoustic scene classification

Performed a literature review of LCA based studies on alternatively fuelled vehicles

Coordinator for the 1st Edition of the ‘Global Business Analytics Challenge’ with a participation of 500+ teams, India’s first internationally hosted case study competition in association with Global Millennium Summit, an initiative of SRCC, Delhi in a triangular partnership with Business Club, IIT Kharagpur and Consult Club, IIM Ahmadabad.Coordinator for the 5th Edition of the ‘Indian Case Challenge’, and coordinated with a team of 24 members to double the participation to 1200+ students compared to the previous edition

I was given the task of meta tagging over 1000 JEE(Joint Entrance Examination) related questions I provided comprehensive solutions to the problems and graded them on a difficulty scale