adithyan js

Data Scientist

• Job Details:Spearheaded the development of an AI-powered hybrid chatbot for Ford's STA engineers. Combined Langchain SQL agent for structured data queries from BigQuery and a retrieval-augmented generation (RAG) agent for unstructured documents like PDFs and PPTs via Elasticsearch. Integrated with Chainlit and deployed on Google Cloud, leading to a 60% faster resolution of supplier quality issues. Directed a multi-model warranty and quality analytics initiative that included anomaly detection, leading/lagging defect indicators, and 3MIS cost prediction. These predictive models reduced warranty claims and quality leaks, generating a projected savings of $150M annually. Rebuilt a deprecated vehicle semiconductor allocation tool originally based on IBM CPLEX using Google OR-Tools. Enabled dynamic multi-constraint optimization under shortage scenarios, achieving a 45% improvement in allocation accuracy. Built the 8502 product – an automated rule-based validation pipeline for upstream supply chain data. Ensured 70% reduction in downstream planning rework and enhanced data reliability across production systems. Designed modular dash UI templates that streamlined visualization for internal tools like OBS-SO, reducing dashboard dev time and improving consistency. Automated end-to-end ingestion-to-reporting data workflows using Apache Airflow via Astronomer DAGs. Orchestrated complex ETL processes across structured and unstructured data. Mentored junior data scientists and served as SME on AI toolchains, LLM integrations, and warranty analytics.

Senior Data Processing Analyst

• Job Details:Led the data integrity function for multiple global FMCG clients, guaranteeing high-quality data outputs through validation, cleansing, and advanced analytics. Engineered a robust OCR + image segmentation system using OpenCV, Caffe, and Pytesseract, allowing automated extraction of text from complex FMCG product images. Conceptualized and delivered an AI-powered workflow manager using Random Forest and Streamlit, reducing operational overhead by 50% through smart automation. Applied EDA, hypothesis testing, and imputation techniques to drive actionable insights from consumer behavior datasets. Developed a geo-matching and client integration framework using Haversine distance logic and fuzzy string matching in Python.

Data Scientist

• Job Details:Directed the 'Pumpkin Tales' project for a hospitality brand. Developed NLP pipelines for sentiment analysis, named entity recognition (NER), and category classification using BERT, XLNet, FastAI, and Transformers. Achieved a 10% uplift in predictive accuracy for Sterling Resorts by implementing IV/WOE feature selection, SMOTE balancing, and XGBoost modeling. Developed multivariate time series models (VAR, MARIMA) and non-linear classifiers to predict steel price fluctuations for investment analytics in the I-Steel project. Delivered data pipelines and ML models deployed on Google Cloud Platform for automated real-time insights delivery.