Introduction
I am an AI Engineer at Commerzbank in London with nearly three years of professional experience. I hold a Master's degree in Artificial Intelligence and a Bachelor's degree in Computer Science. I specialise in agentic AI, RAG pipelines, and MCP servers, building on a strong background in backend and full-stack development. My work focuses on taking AI from prototype to production, automating workflows, and building systems that remain reliable. Outside of work, I build my own projects and contribute to open source.
Education
I earned my Bachelor's degree in Computer Science with First Class Honours from Royal Holloway, University of London. The programme built my foundation in software engineering, databases, algorithms, and machine learning. My final year project was a full-stack social media platform focused on community discussions and engagement.
I then completed my Master's degree in Artificial Intelligence with Distinction at King's College London. I studied machine learning, deep learning, data mining, and computer vision. My dissertation investigated alignment in Large Language Models (LLMs). I developed a novel hybrid-training technique using LoRA that prevents catastrophic forgetting during fine-tuning for hybrid-reasoning models. This allows hybrid-reasoning models to keep their capabilities whilst learning new tasks.
Professional
I currently work at Commerzbank as an AI Engineer. I architect enterprise MCP servers that expose internal platforms, such as ServiceNow and other banking platforms, to autonomous agents. I lead the experimentation and deployment of on-premise LLMs to lower operational costs and enabled higher usage limits across the organisation. I design and implement bespoke RAG pipelines, including Vector RAG, Agentic RAG, and hybrid architectures, tailored to specific team requirements.
As Chair of the London AI Working Group, I discuss and drive the implementation of AI to increase operational efficiency for the London branch. As AI Enabler for the Market Data cluster, I discuss and drive AI adoption across the whole cluster, which spans London, Frankfurt, Prague and Łódź. In both roles, I deliver workshops, training sessions, and presentations, run user surveys to identify pain points, and advise teams on when to use RAG, MCP, or plain deterministic logic.
Previously, I worked as a Full-Stack Software Engineer at the bank. I built backend services with Spring Boot and frontends with Next.js, React, and TypeScript. I delivered the Rates App, giving front-office users instant access to interest rate data from Bloomberg and Refinitiv, and the Application Status dashboard, now used across the bank for real-time monitoring. I developed core microservices for authentication, email, and LDAP permissions used by our web-app, services and Symphony bots. I also re-architected the legacy Symphony bot codebase, which eliminated 80% of maintenance overhead, halved development cycles, and delivered £30,000 in cost savings. On the infrastructure side, I containerised services with Docker and built CI/CD pipelines with TeamCity and SonarQube, cutting deployment times from hours to minutes.
Extras
Open Source
I contribute to the GNOME desktop environment. I helped develop the Quick Settings feature and gave feedback on core applications such as Files, Terminal, and Settings. I also helped implement the details pane for Files UWP on Windows by providing feedback, suggestions and testing. Earlier, I worked as a Student Software Engineer for the Google Developer Student Club at Royal Holloway.
Projects
Applied AI applications. I build full AI-powered products, from a multi-LLM chat client with agentic RAG and tool execution to media generation platforms with payments, quiz generation and workflow automations using flow-based programming.
Agentic tooling and MCP servers. I build tools that let AI agents act on real systems, including an MCP server for controlling the GNOME desktop and one for automating Excel with over 60 tools. I also built a proxy that makes an internal GenAI API OpenAI-compatible.
Local and private AI. I built a VS Code extension that brings local Llama.cpp models into GitHub Copilot for private, low-cost assistance.
LLM research. My dissertation work on hybrid-training with LoRA falls here, alongside experiments fine-tuning and benchmarking open models.
Machine learning fundamentals. I have built a neural network from scratch and a Vision Transformer from scratch to understand the algorithms below the framework level.
Full-stack applications. I build production and learning applications, including a car dealership platform serving a real business, a notes app, a discussion forum, and a real-time messaging app. These were built using various stacks and types of databases (relational and document).