• Led cross-functional teams within the Generative AI CoE, driving innovation and strategic initiatives across the Asia Pacific region
• Led capability assessments and defined solution roadmaps, supporting proposal strategy across multiple client engagements

Client: Rio Tinto

• Led the design and delivery of an Agentic AI-powered IntelliOps platform using agent-based architecture to correct FLOC data, improve BOM accuracy, model component life, predict replacements, and optimise labour, material and shutdown planning

Client: Johnson & Johnson

• Led the design and delivery of a Generative AI platform for automated quiz generation, content evaluation, and knowledge checks using LangChain, FastAPI, AWS, and React, improving training throughput and reducing manual effort
• Acted as Solution Architect and Delivery Lead overseeing end-to-end implementation of generative AI solutions, ensuring alignment with business objectives and technical feasibility

Client: PepsiCo

• Built and deployed granular volume forecasting models across the North American market using Python, PySpark, and Databricks, achieving 90% accuracy and improving planning across customer, location, and package dimensions
• Designed scalable MLOps pipelines in Azure, implementing CI/CD and Continuous Training processes, enhancing reliability, monitoring, and deployment efficiency

Client: Dr Reddy's Laboratories

• Developed a GenAI-based tool to automate the extraction of key data from US Government pre-solicitation and solicitation bids, accelerating bid creation for sales and supply chain teams
• Developed a competitive intelligence pipeline leveraging FDA inspection news via Google News API and web scraping, integrated with PaLM 2 for analysis, and deployed using Airflow, Vertex AI, and BigQuery on Google Cloud Platform

Client: Rolls-Royce

• Explored unsupervised learning methods to predict aircraft engine part failures, led the deployment of physics-informed ML predictive maintenance models for two parts across 1,000+ engines, and built an autoencoder-decoder model to detect anomalies by modelling new engines on the test beds using 250+ parameters
• Created a "Human with AI" tool that saved 1,200+ man-hours by assisting SMEs in diagnosing engine issues
• Developed a supplier optimisation engine that influenced £180M in sourcing decisions

Client: Unilever

• Developed a proof of concept (PoC) application that assigned jobs in a manufacturing plant based on contractor preferences, past performance and availability

Client: OLAM International

• Set up a Data Centre of Excellence using Azure Cloud and built scalable data pipelines

• Collaborated on data science initiatives to optimise store performance and operations across Tesco's UK, Central Europe, and Ireland markets in the Space, Range, Display, Merchandising, and Promotions domain

Client: Walmart International

• Improved under-stocking for Walmart's LATAM supply chain, delivering $1.2M/month in savings by modelling supplier delivery risk (75% accuracy), forecasting lead times (85% SMAPE), and suggesting an optimised EOQ/reorder points that integrate these risks via integer programming and Monte Carlo simulations
• Built an NLP-based PoC to estimate ingredient composition and cost in food items using USDA nutrient data, aiding sourcing managers and food scientists in procurement decisions
• Collaborated with cross-functional teams to develop an interactive web application that empowered sourcing managers and buyers with real-time decision-making tool