Lee Dudek

AI/ML Engineer who has built production-grade agentic and traditional ML based applications for a variety of organisations. I've built systems to identify misinformation, estimate carbon emissions, find funding opportunities and find fraud. I've worked in early stage start ups where I built the product from scratch and more mature companies where I came in to evaluate and improve what exists.

Data Scientist - Grantify - London

(2024-11 - 2026-05)

• Engineered a Grant Recommendation Engine: Built a high-performance matching system to connect users with grants, achieving 98% recall and high-80s precision.
• Developed LLM Evaluation Frameworks: Collaborated with domain experts to create a custom evaluation rubric and a "human-in-the-loop" system to benchmark and spot-check LLM outputs.
• Orchestrated Data Pipelines: Built a comprehensive grant database via multi-source scraping and integrated it with AWS RDS and BigQuery.
• Performance Monitoring: Developed a real-time data dashboard using PostHog and LangFuse to monitor performance drift and alignment between ingested data and user matches.

Machine Learning Engineer - Climatiq - Berlin

(2024-04 - 2024-11)

• Optimised Carbon Emission Mapping: Engineered an "AI Autopilot" using a bi-encoder (embedding) and cross-encoder (classifier) architecture to map corporate product lists to our curated database of human activities, products and services and their associated emissions.
• Advanced Re-ranking: Implemented cosine similarity for initial retrieval followed by a cross-encoder trained on proprietary data for more accurate re-ranking.
• Contextual Data Enrichment: Improved embedding quality for obscure data by concatenating scraped descriptions to entity names and utilising LLMs to synthesise context for material lists.
• Model Deployment: Managed the full ML lifecycle using MLFlow for experiment tracking and AWS SageMaker for production deployment.

Machine Learning Engineer - Whisp - London

(2022-09 - 2024-04)

• Technical Leadership: The primary ML engineer for a blank canvas start up. I designed and built the NLP pipelines and managed a team of 3 engineers following a successful seed round.
• Fact-Checking Infrastructure: Built an automated pipeline for extracting and verifying news claims against a database of trust-weighted sources.
• Cloud Orchestration: Designed the backend architecture using AWS Step Functions to glue multiple ML models into a cohesive verification pipeline.

Python Developer & Comms Exec - Legal & General - Brighton

(2017-06 - 2021-05)

• Process Automation: Transitioned from a senior communications executive role to a technical focus by teaching myself Python.
• Document Search Application: Developed a custom search tool to scan document repositories and our websites for any copy which would be impacted by changes in regulation, product offerings or tax law.

MSc - Artificial Intelligence - Bath University (2021)

BSc - Economics - Brunel University (2011 - 2015)