About Me
Experience
Working as a Platform Engineer @ Fanduel
Worked with an early stage startup to help them establish best practices for model training and deployment. I built an ML/MLOps pipeline prototype in AWS using S3, DVC, Weights & Biases, Terraform & Sagemaker and documented the process. I also worked on fine tuning and deploying a Large Language Model (Alpaca 7B and 13B) on a custom dataset. I benchmarked the models to enable easier decision making on which model to deploy to production.
Worked as an MLOps Engineer where I built efficient data engineering pipelines using DBT to transform raw data from Snowflake into clean data. The pipelines were designed to follow best data engineering practices with observability, reproducibility, testing, and version control. I also built data engineering pipelines to move clean data from Snowflake to a production database that could be used by the model APIs (MongoDB). The pipelines were built to be platform agnostic and could move data from any source to any destination. I built infrastructure using AWS Sagemaker to streamline the model deployment process to production, where each model was thoroughly checked against an unseen test dataset and all metrics were saved to the MLflow model registry for review. I successfully deployed multiple models on APIs using Docker, FastAPI, AWS Lambda, API Gateway, Elastic Beanstalk, and Terraform. I set up a monitoring server using Grafana to track internal infrastructure performance, data pipeline execution, and data distribution in production.
I built deep learning models to track construction and road construction activity from satellite imagery using PyTorch, PyTorch lightning, Python, and Docker. I also set up infrastructure to track Machine Learning experiments using MLflow. I deployed the deep learning models to AWS batch to run offline predictions, and all model repositories were deployed using CI/CD with Terraform and AWS CodePipeline. Additionally, I built front-end applications using React and Dash to showcase the models to non-technical stakeholders in an engaging and interactive way.
I developed an unsupervised learning model using NLP embeddings with scikit learn to detect participants who answered screener questions inconsistently. To deploy this model I built a web interface using Flask with React & Redux to enable the Q&A to interact with the model to detect fraudulent users.
During this period of time I travelled to Scotland where I worked full time as a chef and a lifeguard while learned programming in my spare time. Through a series of online courses and personal projects I learned front-end development and progressed to machine learning & deep learning. During this time I also participated in the Mongol Rally, the longest charity rally in the world. We raised £1k for charity APIC Pandora.
As part of the sales team, I was involved in the growth of the spanish branch of the company. My tasks included: Prospecting clients to create new sale opportunities; Keeping track of current and potential clients; Delivering succesful sales.
Education
Learned AI foundations, data mining and machine learning techniques. I also used applied statistics to solve business problems. I worked in teams to complete projects including using swarm algorithms (PSO) to optimise the parameters of a neural network and building a conversational agent to assist older adults. Finally, I completed my thesis as part of an industrial placement in UserTesting
Learned the foundations of business creation & development, marketing, finance, accounting and law. During this time I participated in the Erasmus+ program studying in Düsseldorf for an academic year.
Project Showcase
Rebuilt my reinforcement learning project to be up to date
A voice assistant designed to assist eldery people
An agent learns to play space invaders from image inputs.
A telegram bot that (almost) speaks like Don Quijote de la Mancha.
Participated in the Mongol Rally where we raised £1k for charity.