Karan Gadodia

Machine Learning Engineer / Generative AI Engineer - PureFrame Labs - Maharashtra, India

(2026-01 - 2026-04)

Architected and deployed an enterprise-grade Retrieval-Augmented Generation (RAG) chatbot using Qwen 2.5, FAISS vector database, embeddings, and prompt engineering techniques on large-scale Excel knowledge bases, reducing manual lookup effort by 70%.
Built scalable Python ETL and web scraping pipelines for automated extraction, cleaning, chunking, embedding generation, and semantic indexing of structured government-portal datasets for LLM-powered search and retrieval.
Implemented semantic similarity search and grounded retrieval pipelines to reduce hallucination and improve contextual accuracy of LLM responses.
Conducted proof-of-concept demonstrations and technical walkthroughs for business stakeholders, translating AI capabilities into practical business solutions.
Collaborated with cross-functional teams to gather requirements, design solution architectures, and present implementation approaches for enterprise AI initiatives.

Data Scientist - Abhay Industries Solar Division - Maharashtra, India

(2024-09 - 2025-09)

Built a 24-hour solar generation forecasting model using XGBoost across a 40MW+ fleet of 15+ sites, achieving 6.8% MAPE and outperforming ARIMA and Prophet baselines.
Developed a predictive maintenance classifier using Random Forest on 3 years of inverter sensor data, detecting equipment failure signals 7–14 days in advance with 87% accuracy.
Designed a soiling loss prediction model using pollution, rainfall, and wind datasets to optimize panel-cleaning schedules, reducing generation loss by 11%.
Deployed ML models into a Streamlit-based fleet intelligence dashboard with automated alerts, reducing operations response time from 3 days to same-day resolution.

Data Analyst / Data Science - Shegaon Shree Agrasen Co-op Bank - Maharashtra, India

(2024-05 - 2024-08)

Performed exploratory data analysis on 10 years of banking data covering deposits, loans, NPAs, and profitability trends across 14 branches using Python and Power BI.
Conducted NPA risk and credit risk correlation analysis identifying drivers behind Gross NPA reduction from 13.29% to 9.28%, supporting strategic management decisions.
Built a 3-year financial forecasting model using Facebook Prophet to project deposits, loan growth, and profitability for board-level planning.
Designed interactive Power BI dashboards visualizing branch-wise KPIs, NPA movement, budget variance, and financial performance metrics.
Automated monthly branch performance reporting using Python and openpyxl, eliminating 3 days of manual reporting effort every month.

Associate Software Engineer → Associate Data Scientist → Developer I - UST - Pune, Maharashtra

(2022-07 - 2024-04)

Built supervised and unsupervised machine learning models for customer segmentation, churn prediction, and predictive analytics using Logistic Regression and Random Forest algorithms.
Designed end-to-end ML pipelines covering data ingestion, ETL, feature engineering, model training, monitoring, and deployment using Python and Scikit-learn.
Improved predictive model performance through hyperparameter tuning, cross-validation, and feature engineering techniques.
Developed a Streamlit-based churn prediction application enabling non-technical business users to leverage ML insights for customer retention strategies.
Created Power BI dashboards visualizing KPIs, revenue trends, customer analytics, and inventory insights for cross-functional stakeholders.
Worked directly with business stakeholders to understand requirements, translate business problems into machine learning solutions, and present analytical findings through workshops and demonstrations.
Supported solution design activities, effort estimation, and proof-of-concept development for AI and analytics initiatives within Agile delivery teams.

About