Art Eidukas

Built, optimized, deployed and maintained text-to-speech serving systems at Spotify, supporting products such as Spotify DJ and other voice-driven experiences used by millions of listeners.

Worked deep in the inference stack: CUDA kernel profiling, PyTorch model optimisation, writing an ML serving stack in Python.

Integrated their technology into the Spotify stack post-acquisition.

Over my tenure in ONS I have participated in countless projects and initiatives to improve the way the UK does statistics.

I went from an economist to a data engineer, to a data scientist, to a software engineer focused on productionising research code.

• Built computer vision and NLP pipelines.
• Improved inflation statistics.
• Learned about working in technologically restricted environments.
• Wrote distributed data processing pipelines for large datasets
• Worked on the COVID19 pandemic response.
• Mentored in government-wide Data Science programs.

This is where I grew up professionally and developed a strong value for money mentality: build what matters, spend carefully, and make systems useful under real constraints.