David J. Email & Phone Number

I am a Data Scientist at Airbus working in the Airline Sciences Advanced Projects team. My areas of expertise are deep learning, computer vision, Python software development and AWS.After joining Airbus in 2021, I was part of the Information Management Graduate Scheme in Filton, doing placements within teams working on AI/ML topics.At Airbus, I have worked on a range of projects including:- Explainability of state of the art object detection models such as YOLOv5/8 and other in-house model architectures - this includes TensorFlow, PyTorch and ONNX models.- Adversarial training of models using GAN-generated images and using explainability results to target model improvement during training as well as part of fine-tuning.- Helped build a time-series anomaly detection Python library using software engineering best practices - this included setting up the CI/CD Jenkins pipeline, writing unit and integration tests and using libraries such as Pydantic, MyPy, PyLint, PyTest.- Building, training, deploying models on AWS SageMaker (Studio/JupyterLab and Code Editor) - this includes writing dataloaders to pull data from S3, setting up training jobs and deploying SageMaker endpoints.- Establishing AI/ML best practices on AWS for my team, including creating custom Data Science EC2 AMIs, launch templates, SageMaker Studio + Code Editor, MLflow tracking servers, building a utility Python library for working with data from other AWS services from SageMaker and EC2 and enabling colleagues to migrate to EMR Serverless.- Defining and setting up the architecture on AWS for running studies at scale using Lambda functions and Athena. This consists of a pipeline including study configuration, launching computations, retrieving results, extracting features for further analysis and visualisation through Grafana.- Built a proof-of-concept live unsupervised anomaly detection demonstrator using Microsoft Flight Simulator - this was written in Rust and Python and used Grafana and InfluxDB to stream data, classify it and visualise it.Outside of work, I developed a Python library for explaining neural networks using the Tinygrad deep learning framework. This was inspired by the popular Xplique and Captum libraries for TensorFlow and PyTorch, respectively.Other experience includes:- Git- GCP - completed a week-long Google course on Vertex AI- Linux- DockerPrior to joining Airbus, my Master's thesis on deforestation detection using semantic segmentation of satellite imagery was published as a peer-reviewed paper.