Varishth Bhaskar

About Varishth Bhaskar

As a Data Scientist and Machine Learning Engineer with over 3 years of experience, I…

Experience

1. Associate Data Scientist

   Nits Solutions

   Aug 2024 — Present · Gurugram, IN

   Key Projects & Responsibilities: UOM ID Prediction using Large Language Models (LLMs)Fine-tuned a pre-trained LLM (Gemini-2.0-Flash) using prompt engineering for Universal Operation Code (UOM ID) prediction across 192 Repair Order IDs.Applied NLP techniques and regex-based text preprocessing to clean and standardize service descriptions.Achieved over 90% prediction accuracy in the ongoing development pipeline, with a focus on model generalization and deployment optimization. Time Series Forecasting of Labor HoursBuilt a predictive model to forecast labor hours over a 6-month window for 100+ automotive dealerships.Engineered features including calendar attributes, holidays, and moving averages to capture short-term and seasonal trends.Trained dealership-specific forecasting models using Facebook Prophet and AutoGluon, achieving <10% mean error in over 70% of cases. OP Code Harmonization via BERT-Based ClassificationDesigned and trained two distinct BERT-based classification models to identify UOM IDs for maintenance vs. non-maintenance tasks.Preprocessed operational descriptions using NLP pipelines and regular expressions to enhance input quality.Reached 93% accuracy for the maintenance model and 96% accuracy for the non-maintenance model, enabling streamlined UOM mapping across systems.Tools & Technologies:Python, BERT, LLMs (Gemini-2.0-Flash), Prompt Engineering, AutoGluon, Prophet, Time Series Analysis, Text Preprocessing, Regex, Scikit-learn, NLP, AWS (S3, SageMaker), Git

Education

• Jamia Hamdard

  Bachelor of Technology - BTech

  2019 — 2023