Tim Staley Email & Phone Number

Uk

My focus at Exscientia has been driving and supporting robust software engineering practice amongst our research scientist teams, and building pathways to allow rapid deployment to production.* Software lead on a complex workflow system migration, from an external platform to an internal 'green field' implementation. Under extremely tight time constraints we achieved a seamless migration between workflow systems, successfully adopting a new workflow definition framework internally… Show more My focus at Exscientia has been driving and supporting robust software engineering practice amongst our research scientist teams, and building pathways to allow rapid deployment to production.* Software lead on a complex workflow system migration, from an external platform to an internal 'green field' implementation. Under extremely tight time constraints we achieved a seamless migration between workflow systems, successfully adopting a new workflow definition framework internally. Provided cost savings of £1.5m+ p.a. and ensured long-term viability through use of 'off the shelf' open-source technologies. * Built a community of practice around Python and reproducible deployments together with a well supported project framework to streamline adoption. Reduced project setup time from two days to ten minutes. These standards were adopted company wide, improving consistency and maturity of engineering practice across multiple teams and functions. Reduced project startup time from 3 days to 30 minutes.* Hired an RSE team to provide both infrastructure and direct support for the AI research group. Devised job specs, set up an interview process including a software skills test, and ran ~40 interviews across 3 months. Prioritised and planned infrastructure and direct support timelines to help research groups deliver on their quarterly goals.* Identified, documented and stabilised ad hoc ML deployments used on a daily basis, before implementing auditable IaC deployments. Profiled and debugged problematic model wrapper code. Query failure rate reduced by two orders of magnitude while throughput increased 10x.* Tech: Argo Workflows & Hera, Kubernetes / AWS EKS, Terraform, Argo CD. Prefect, Dask & AWS Batch. MLFlow.Python packaging, linting & testing ecosystem (mypy/ruff/pytest/nox), Cookiecutter tooling. Docker, PR driven CI/CD. Pydantic & FastAPI. Show less

Oxford, England, United Kingdom

Oxford, United Kingdom

Mind Foundry is a B2B SaaS company with an early focus on AutoML - i.e. bringing together a blend of automated and user driven feature engineering, automated model selection, hyperparameter optimization, evaluation and deployment tooling.My role at Mind Foundry was a blend of machine learning engineer, data scientist, and some algorithmic research. In practice this meant a mixture of core-product software engineering, 'research spikes' to assess feasibility and derisk new features, and… Show more Mind Foundry is a B2B SaaS company with an early focus on AutoML - i.e. bringing together a blend of automated and user driven feature engineering, automated model selection, hyperparameter optimization, evaluation and deployment tooling.My role at Mind Foundry was a blend of machine learning engineer, data scientist, and some algorithmic research. In practice this meant a mixture of core-product software engineering, 'research spikes' to assess feasibility and derisk new features, and in-house algorithmic library development.* Lead and delivered multiple data-science consulting projects. Common themes of large data processing and sampling, dimensionality reduction techniques, and visualisation to provide insights into global structure of the datasets. One such project (processing over 1B rows of time series data) was key to securing the client as a lead investor.* Explored and evaluated various interpretability methods. Wrote performant implementations for the main product (feature importance, SHAP).* Rewrote sampling and stratification implementation for our platform after identifying issues around imbalanced classification datasets. Improved benchmark score from 91% to 95%.* Implemented a time series forecasting framework that became a spin-off product.* Tech: Pandas, Numpy, Scikit-Learn, UMAP, HDBScan, Statsmodels. Flask. Dask, Snakemake. AWS EC2. Show less

Oxford, United Kingdom

For around 18 months I took on consulting work as an independent scientific / research software engineer; clients included the Square Kilometre Array (an international radio telescope project with significant international funding and extremely demanding software requirements), a small academic spin-out / start-up, and smaller maintenance engagements on open-source academic code-bases.Sample engagements included:* A reference implementation of radio astronomy algorithms from the… Show more For around 18 months I took on consulting work as an independent scientific / research software engineer; clients included the Square Kilometre Array (an international radio telescope project with significant international funding and extremely demanding software requirements), a small academic spin-out / start-up, and smaller maintenance engagements on open-source academic code-bases.Sample engagements included:* A reference implementation of radio astronomy algorithms from the literature, for cross-validation and instruction of externally developed C++ code; combined with coordination and direction of remote C++ team.* Rapid prototyping of a SQL/JSON hybrid data-backend for a mobile application.* Forensic debugging and reconfiguration of legacy web-server systems.* Tech: Numpy, Django, Postgres, Celery, LAMP stack, Ansible. Virtualbox & Vagrant. Show less

Oxford

This role was effectively that of research software engineer, embedded within a software-intensive radio-astronomy research group. The focus was a mixture of investigating novel algorithmic approaches to automated transient discovery and alerting, and supporting colleagues with challenging data-analysis and processing workflows (e.g. working with 100TB datasets from the LOFAR telescope).I was lucky enough to join a project funded via a 5-year ERC grant, which enabled long-term… Show more This role was effectively that of research software engineer, embedded within a software-intensive radio-astronomy research group. The focus was a mixture of investigating novel algorithmic approaches to automated transient discovery and alerting, and supporting colleagues with challenging data-analysis and processing workflows (e.g. working with 100TB datasets from the LOFAR telescope).I was lucky enough to join a project funded via a 5-year ERC grant, which enabled long-term collaboration with a distributed team of developers working on a large software project (the LOFAR transients pipeline, 'TraP'), concurrently with development of a novel software-ecosystem for responsive transient discovery and alerting (the VOEvent / 4PiSky alerts database, the ALARRM automated data-reduction pipeline). I was lead and co-author on several peer-reviewed publications during this time - see personal website for details.Highlights:* Development of a modular open source ecosystem for automated transient astronomy, enabling rapid notification and automatic scheduling of followup observations.* Core developer on a transients detection pipeline, still under active use and development.* Prototyping of a probabilistic decision making system for prioritising follow up of events.* Numerous publications including a Science paper (A radio jet from the optical and x-ray bright stellar tidal disruption flare ASASSN-14li, 2016).* Tech: Django, Postgres/MonetDB, PyMC3, Twisted (async), LXML, pexpect. Gaussian Process models and Monte Carlo simulations. Show less

During my PhD I worked in the lucky-imaging group at the Institute of Astronomy, developing novel, high performance data-reduction techniques for a high-speed low-noise optical-imaging instrument. As a result we achieved images with resolution and sensitivity comparable to the Hubble space telescope, for a tiny fraction of the cost.

Ph.D., Astronomy

Supervisor: Prof. Craig Mackay Thesis title: Lucky Imaging - Beyond Binary Stars

M. Sci, Astrophysics

Supervisor: Dr. Ian Parry Dissertation title: Dual Plane Imaging Project entailed simulations of novel instrumental optics design.

B.A, Mathematics

Main focus on applied mathematics, specifically fluid dynamics and wave theory.