Headline

AI Engineer | GenAI • LLMs • Agentic AI | Ex-BAJAJ, TATA | IIT Kharagpur’21 | AWS | RegTech | InsurTech | MusicTech

About

At CUBE Global, my role as Senior Data Scientist involves leveraging expertise in Artificial Intelligence, Generative AI, NLP and Machine Learning to build data-driven solutions for complex challenges in the RegTech domain. This position builds upon my experience in BFSI/InsurTech, where I crafted models that significantly enhanced retention rates and generated substantial additional gross written premium. With a solid foundation from IIT Kharagpur in Exploration Geophysics, I apply rigorous analytical and problem-solving skills to complex business challenges. My core competencies include Machine Learning, NLP, Generative AI, LLMs, RAG and AI/ML systems, complemented by my recognition as Analytics Champion and Champion of Champions at Tata AIG. I’m passionate about continuously pushing the boundaries of data science and AI to create innovative solutions that drive business impact, improve customer experiences, and enhance operational efficiency.

Location

Bengaluru

AI stack

claude code

gemini

hugging face

langchain

langgraph

openai

Work experience

Senior Data Scientist at Cube, Sep 2026 - Present · 2 mos, Bengaluru

Associate Senior Manager Analytics at Bajaj General Insurance, Nov 2023 - Aug 2026 · 2 yrs 10 mos, Pune District

Data Scientist at Tata AIG General Insurance Company Limited, Jun 2021 - Nov 2023 · 2 yrs 6 mos, Mumbai, Maharashtra, India, • Build private car customers renewal propensity model, increasing retention rate by 4%. Delivered additional GWP of 125+ Cr (in 2.5yrs) • Build ML model for Private car risk based customer level pricing model in order to increase revenue and maximize the retention of desirable customers on the basis of renewal and claim propensity • Delivered GWP impact of 16lac extra revenue at one location with ₹1300 extra premium on every policy ( in 4 months). And increased retention by 5% for Bangalore and 10% for other 5 locations • Created call volume forecasting model and optimized call center executive staffing using Amazon DeepAR forecasting algorithm. Impacting 30% increase in SLA • Implemented and learned time series models like SARIMAX, Prophet and auto time series models like AutoTS, DeepAR and FEDOT • Experience in end to end Model Development – right from Data Engineering (using PostgreSQL in DBeaver Data Lake), Cleansing, EDA, Feature Engineering and Modelling

Machine Learning and Artificial Intelligence at Ethical Edufabrica Pvt. Ltd, Aug 2020 - Sep 2020 · 2 mos, New Delhi, Delhi, India, •Predicted Stock Market Price of 20 years of ONGC stock record using keras stacked Long Short Term Memory (LSTM) neural network layer •Preprocessed the dataset by sorting according to date, used MinMaxScaler for scaling and considered closing price as target variable •Applied Dense, LSTM, Dropout keras layers, used Adam optimizer, mean squared error as loss function with 50 epoch and 60 batch size •Achieved Root Mean Square Error (RMSE) equal to 9.41 and evaluated the model with R squared score metrics and got 0.95

ML Training at Ethical Edufabrica Pvt. Ltd, Aug 2020 - Sep 2020 · 2 mos, •Calculated top 10 features of mobile price Kaggle dataset using correlation matrix, obtained 89.5% and 95% accuracy using DTC and SVM •Predicted 3 flower species from 4 features of Iris Dataset using K-means clustering and Random Forest Classifier with an accuracy of 97% •Applied FFNN on MNIST data with sparse categorical crossentropy loss function ,ReLU ,softmax activation function with accuracy of 92% •Generated Spam Mail Classifier by NLTK text processing on ~5000 Kaggle dataset, with an accuracy of 98% using Logistic Regression

Summer Intern at Indian Institute of Technology, Kharagpur, Jun 2020 - Aug 2020 · 3 mos, Kharagpur, West Bengal, India, •Generated Forward Model and calculated Apparent Resistivity and Phase using Resistivity and Thickness of 1D - MT Data in MATLAB •Formulated Inversion Model using SVD technique to invert Jacobian Matrix containing Resistivity and Thickness values of 1D-MT Data •Obtained Root Mean Square Error for resistivity in the range of 0.005 - 0.8 and thickness in the range of 0.001 - 0.04 on noise free data •Generated Synthetic data from given true model using forward model and added 5% of random noise on synthetic data using MS Excel •Calculated Root Mean Square Error for resistivity in the range of 0.78 - 1.15 and for thickness in the range of 1.59 - 1.96 on noisy data

Education

Bachelor's and Master's degree, Engineering, Indian Institute of Technology, Kharagpur

Public links

github: https://github.com/rohitkr99

rohitkr99.github.io: https://rohitkr99.github.io/portfolio/