Work
Where technology meets strategy: AI and data for banking and finance today, production AI for manufacturers before that.
Metyis
Leading AI and data development for banking and finance.
- Associate Data Science Consultant2025 to now
Accenture Applied Intelligence
Building production AI and machine learning for supply chain manufacturers across EMEA and the US, from GenAI agents to computer vision.
- Context
- Seventeen production ML solutions had to move to a private cloud environment.
- What I did
- Planned and coordinated the migration across EMEA and the US, with daily stand-ups and weekly stakeholder meetings. Built CI/CD pipelines and served models on Azure Kubernetes Service.
- Result
- 17 end-to-end migrations delivered in four months.
- Context
- US factories needed to optimise how chocolate is sprayed onto peanuts.
- What I did
- Machine learning models for the spraying process, an ETL pipeline for production data, computer vision for real-time monitoring, and CI/CD from training to deployment.
- Result
- A scalable ML solution with real-time monitoring and quality assurance on the line.
- Context
- A four-week RFP to detect, predict and correct issues in production plants.
- What I did
- Forecasting models and mathematical modelling algorithms. Python, Databricks, PySpark.
- Result
- Production yield raised to 99.1%, delivered within four weeks.
- Context
- Efficiency bottlenecks in production were causing financial losses that were hard to trace.
- What I did
- Designed a track-and-trace algorithm and a financial model, deployed with CI/CD at 98.5% automated test coverage, with Power BI dashboards for the plant teams.
- Result
- Material overuse down 15%, and the dashboards adopted by six US plants.
- Context
- A mature ML project needed better efficiency and model performance.
- What I did
- Refactored the codebase for constant-time lookups, retrained the models with XGBoost, added robustness logic and built a data-quality module with the team.
- Result
- Higher R² scores and a faster, more robust codebase.
- Context
- Plant teams needed GenAI agents that could work with data, documents, images and audio.
- What I did
- Built a modular platform (frontend, gateway, router and SDK) on FastAPI and LangChain, with predictive-maintenance models inside the agents, deployed on Azure to Teams and Slack.
- Result
- Up to 30% fewer production stoppages, 81% of failures detected early, and 3,000+ concurrent users.
Education
- Master's in AI and Analytics
- Electronic Engineering and Business ManagementKing's College London