Headline

Data Scientist III @ Walmart Global Tech|CSE BTech NIT DGP| 7 YoE|AIML|Deep Learning|NLP|LLM|GenAI|Advanced Analytics

About

As a Data Scientist III at Walmart Global Tech India since September 2024, my work focuses on developing personalized substitute recommendation systems using ranking models and GenAI-based evaluation frameworks. My academic foundation includes a Bachelor of Technology in Computer Science and Engineering from the National Institute of Technology Durgapur. My professional experience spans diverse domains, including e-commerce, telecom, healthcare, manufacturing. I have contributed to projects involving GenAI, LangChain, and personalization, with a strong focus on NLP, LLMs, and machine learning solutions. My commitment lies in leveraging technology to drive innovation and deliver impactful solutions.

Location

Bengaluru

AI stack

gemini

github copilot

hugging face

langchain

langgraph

ollama

Work experience

Data Scientist III at Walmart Global Tech India, Sep 2024 - Present · 2 yrs 1 mo, Bengaluru

Data Scientist at Fractal, Dec 2022 - Aug 2024 · 1 yr 9 mos, Bengaluru, Improving user experience with customer support care center by reducing the number of transfer/call back of chats/calls or repeat contact made by customer by analyzing chat and call transcripts. Developed an unsupervised NLP pipeline to identify the top issues for chat/call transfer problem by clustering the data using HDBSCAN after generating embedding from sentence transformer and finding relevant topic from each cluster by using BERTopic modelling. LLMs are used to obtain relevant label for each topic by doing instruction finetuning using zero-shot, few-shot learning. To tackle clustering of vast amount of data, pipelines are built based on sample data and after that developed supervised multi-class classifier model by finetuning prebuilt BERT based model through changing the last layer architecture of the model to incorporate the number of classes and adjust the weights of different layers based on our customized clustered sample data with LLM generated label and tagged rest of the data through prediction of the finetunned model.

Senior Analyst - Data Science at Tiger Analytics, Oct 2021 - Nov 2022 · 1 yr 2 mos, Bangalore Urban, Karnataka, India, Developed NLP pipeline in Pyspark to map ratings and reviews given to different products of the company through the different ecommerce websites to themes by using several NLP techniques like direct matching, ngram matching, prefix suffix matching and pos tagging. Developed sentiment analysis pipeline to map review sentences to positive, negative, neutral sentiments via Azure text analytics API model and built custom LSTM model to improve the accuracy of the API model over 90%. Developed analytical model for sales apportioning among growth drivers for Market Development Organization by considering equal contribution of the GDs and also the variable contribution based on the coefficients of the growth driver variables from regression model and by also applying different business rules on top of the model prediction to meet the business requirements.

Data Scientist at Abzooba, Aug 2021 - Sep 2021 · 2 mos, Kolkata, West Bengal, India, Comparative analysis among the several open-source ML platforms like Dataiku's DSS and Azure databricks to analyze different AI/ML lifecycle features by implementing some dummy ML pipeline on their both local and enterprise-based server.

Jr Data Scientist at Abzooba, Nov 2020 - Jul 2021 · 9 mos, Kolkata, West Bengal, India, Developed ML model through SensiML platform that takes time-series sensor data from manufacturing equipment and predicts a set of possible gestures as normal/anomaly. This model enables to raise an alarm to the plant supervisor for the upcoming machine downtime and perform proper maintenance of the machines.

Software Engineer in Advanced Analytics at Abzooba, Jun 2019 - Oct 2020 · 1 yr 5 mos, Kolkata Area, India, Developed a framework to identify the patients who are at high risk of getting infected by COVID-19 by analyzing and by doing comparative studies among different available data based on patient's past medical history, demographic data, etc. This model helps the care management nurses to get an actionable list of members and can take proper intervention plans. Developed an end-to-end Machine ML model in PySpark starting from data ingestion to evaluation metric generation for selection of the best model to predict patient readmission. This model would help to identify the patients who have a high risk of being readmitted and thus plan for proper intervention for them to avoid readmission. Achieved 73% AUC_ROC score.

Final year Project Mentee at National Institute of Technology Durgapur, Jan 2019 - May 2019 · 5 mos, Durgapur, West Bengal, India, Air pollution prediction of an area of concern by collecting the pollution data from the gas sensor node deployed on the different transportation system. The virtual vehicular sensor network is constructed in such a way that energy consumption of the whole networks while transferring the collected pollution data to the central server will be minimised. The prediction on the dataset collected from the central server is done by using ANN by using deep learning libraries like keras, tensorflow and also back propagation algorithm is used in the feedforward ANN to minimise the prediction error.

Education

Bachelor of Technology, Computer Science and Engineering, National Institute of Technology Durgapur

12th Board, Mathematics and science, Aligunj Rishi Rajnarayan Balika Vidyalaya