Principal (Data Scientist/Data Engineer/Gen AI Engineer)

Architected an agentic, real-time offer orchestration engine where specialized agents (retrieval, pricing/optimization, and policy/guardrails) coordinate to personalise promotions across customer touchpoints. The system uses RAG with FAISS over Delta Lake and low-latency Databricks Model Serving to support on-the-fly decisions and partner integrations; collaborated with PMs and commercial stakeholders to shape the roadmap and evaluate emerging agent patterns for production.

Designed an agent-based data quality service that orchestrates schema detection, entity normalization, and validator/exception-handling agents to clean multi-retailer SKU feeds at scale. Wrapped model calls in PySpark UDFs for distributed inference, automated via Databricks Workflows and CI/CD, enabling near-real-time readiness for downstream agent decisioning in supply and catalog processes.

Developed a multimodal, agentic extraction pipeline where vision, parsing, and compliance agents collaborate to derive brand, packaging, and volume from scanned images using Claude 3 Sonnet with Swin Transformer encoders. Asynchronously orchestrated via Azure Event Hub with enriched outputs persisted to Delta Lake for consumption by search, recommendations, and downstream operational agents.

Implemented a GS1 taxonomy classification service built around cooperating agents for inference, drift monitoring, and auto-retraining governance. Falcon 180B (LoRA-tuned) powers the classifier; a batch pipeline on Databricks triggers model refreshes when accuracy dips below thresholds, supporting reliable merchandising analytics and agent-driven discovery.

Created a hybrid agent workflow where a retrieval agent surfaces candidate matches via embeddings and a reasoning/verification agent (Mixtral 8x7B) adjudicates final receipt-to-SKU alignment. Integrated into a streaming Databricks pipeline to support near-real-time sales operations and exception handling across customer touchpoints.

Built a multimodal attribute inference pipeline structured as cooperating vision-language, rules/consistency, and compliance agents to fill NutriScore, nutrition fields, and packaging types from names and images using LLaMA 3-8B with CLIP embeddings. Designed for fast feedback loops so downstream agents can trust and act on enriched product records.

Developed a GenAI-powered orchestration system that ingests recipes from multiple websites, parses ingredients through structured extraction agents, and dynamically links them to real-time retailer offers via tagging agents. Designed a multi-agent workflow where retrieval agents identify candidate offers, semantic reasoning agents validate ingredient–offer matches, and business-rule agents ensure compliance. Integrated into customer touchpoints so that users could directly click on matched offers, driving incremental sales and partner revenue share.

Developed and deployed a Random Forest model on AWS SageMaker to calibrate IoT device temperatures in greenhouses. Leveraged physics-based features, achieving a 20% improvement in prediction accuracy for real-time monitoring across 8000 devices.

Designed an AI-powered inventory system utilizing time series analysis to predict demand, automating replenishment processes. Improved stock accuracy by 30%, reducing inefficiencies and aligning inventory with dynamic retail needs.

Developed an AI-driven quality control system using computer vision to detect defects, achieving 98% accuracy. Reduced production errors by 35%, ensuring compliance with manufacturing standards and boosting efficiency.

Designed a content-based recommendation system to provide personalized banking product suggestions, increasing product sales by 20% and improving customer satisfaction through data-driven insights.

Developed an interactive chatbot for financial product recommendations using IBM Watson Assistant, improving customer engagement by 30% with a user-friendly interface hosted on a website.

Built a predictive scorecard using logistic regression and lift charts to identify high-probability loan customers, boosting loan acquisition rates by 25% through targeted marketing.

Forecasted monthly sales using advanced time series models, including ARIMA and LSTM, delivering 95% prediction accuracy to enhance dealership-level strategic planning.

Predicted individual medical costs using Ridge, Lasso, and Elastic Net regression models, achieving a 15% improvement in cost estimation accuracy for policy pricing and outreach strategies.

Deployed AI-driven models to forecast demand and optimize inventory, reducing operational costs by 25% and improving logistics efficiency for supply chain operations.

Created an AI-powered chatbot with GPT-based NLP for efficient query handling, reducing response times by 40% and enhancing customer experience on e-commerce platforms.