7+ years delivering end-to-end ML and Generative AI solutions across enterprise environments. Specializing in Agentic AI, RAG pipelines, and demand forecasting.
Enterprise AI systems, open-source research, and published writing — across ML, GenAI, and data engineering.
Linear programming-based engine generating optimized truck shipments against complex supply chain constraints — minimizing cost while maximizing load efficiency. Deployed in production for a large US client.
Consolidated 5 legacy systems (SQL, Elasticsearch, REST APIs, SharePoint) into one conversational AI interface using LangGraph. Reduced candidate identification from 1 week to near-instant for 200+ users.
Automated review of 100+ Excel templates using LLMs, eliminating manual validation entirely. Transformed analyst capacity and boosted compliance accuracy from ~65% to 95%.
AI platform automating end-to-end newsletter creation — content curation, summarization, and formatting — replacing a fully manual editorial process across multiple groups.
Combined depletion forecasting with stock-out probability classification using 5 years of historical data, enabling proactive inventory management and reducing revenue loss from stock-outs.
Regression-based ML models trained on 5 years of historical data to predict roll end dates, improving customer confidence and operational stability for manufacturing planning.
Multiple production RAG chatbots with Natural Language Query capabilities using hybrid RAG architectures. LangGraph for agent orchestration, LangFuse for observability and performance monitoring.
Naïve Bayes text classification model automating ticket categorisation in the internal Incident Management tool across 9 teams. Increased billing frequency from monthly to bi-weekly.
Real-time license plate detection on live highway cameras using OpenCV for vehicle detection and a Tesseract OCR model fine-tuned for UK number plates. Ported and deployed directly onto edge camera hardware.
Campus restaurant chatbot using intent recognition, NER, and a dialogue flow manager. Built with SpaCy and an LSTM-based architecture before the LLM era.
View on GitHubComparative study of ML models across two datasets with contrasting properties — examining generalization, bias-variance tradeoffs, and metric sensitivity.
View on GitHubMonte Carlo stock risk analysis across AWS and GCP. GCP hosts the Flask app; AWS Lambda & EC2 handle compute; S3 manages storage.
View on GitHubBFS/DFS pathfinding for the London Underground network using Prolog. Explores constraint-based reasoning and relational knowledge representation.
View on GitHubAccuracy is only one part of a big puzzle. A clear breakdown of Accuracy, Precision, and Recall — and when each actually matters in practice.
Read on LinkedInA data-driven investigation into global population trends. Digging into the numbers behind a question everyone is asking — with surprising findings.
Read on LinkedIn