- Developing a classification model which predicts UTBMS litigation codes of invoice line items using LLM’s - Developed UIUX using stream lit framework to get feedbacks from users which helps in fine tuning the model- Developing customer legal risk forecaster model using vehicles repair history etc - Developed Dashboards for Insurance team

-Developed an optimized customer email summarization system using the LangChain framework, leveraging advanced natural language processing (NLP) models to efficiently summarize large email conversations.

• One of the Research Assistants in – Yohe’s Lab • Deployed an end-to-end pipeline to classify bat’s diet using MRI scan images of bat-guts.The model is trained using deeplearning techniques. Pre-trained models were also used.• Working on to build a Classification model to predict bat’s diet based on sound and head images of bats.• Dockerized developed models and deployed them in production.Trained ML models were tracked down using MLOps.

• Developed multilingual experience prediction model using LSTM neural networks with given title, description as inputs. Developed Multilabel binary classification model to get important sentences from Job descriptions using Bert.• Created a factory of ML techniques and deep neural networks, to predict important sentences from job-descriptions, as a part of 4 member team.Dockerized developed models to be moved into production.• Designed an end-to-end pipline python service to use the ML models to extract the entities from job descriptions and store the entities into the database(MongoDb , Elastic search).• Developed a model which generates teaser from job description, which helps the customers to get a brief overview of the job without opening the job page and overview of job in Email recommendations.• Optimised the compute time to extract entities from job-descriptions using parallel processing.• Designed multilingual sentence-classifier binary classifier model using XLMR-BERT embeddings and achieved decentprecision and recall values.