Michael Ball Email & Phone Number

Berlin, De

Berlin, Germany

Berlin, Germany

I managed all of Oculid's production infrastructure, machine learning engineering, and software (excl. mobile). Some of the things I've done so far are:- Designed, implemented, and maintained all production infrastructure (AWS, Terraform, Docker, Sentry, Gitlab CI, etc.)- Led the company's transition from notebook hacking to serving models in production: data versioning w. DVC, experiment tracking & CI/CD w. MLFlow, and continuous testing and monitoring of production models with custom tools- Built a standalone data pipeline for automated data collection (enabling Oculid to collect the largest gaze prediction dataset in the world)- Designed and implemented GDPR processes in line with industry best practices.- Guided software engineering team growth: defined roles, interviewed candidates, distributed tasks, and provided mentorship

Berlin, Germany

- Built the repository used by entire data science team for preprocessing, training, fine-tuning, & evaluation of all models.- Achieved 8x speedup in serving predictions through optimisation of model server & refactoring of data preprocessing pipeline- Built a GPU scheduling & prioritisation service to optimise data science job management across our in-house hardware stack.- Redesigned our image feature extractor using a variety of recent CNN innovations; decreased model error by 9%.- Oversaw three month research project into edge computing, TFLite, & serving optimized models on mobile devices.

Auckland, New Zealand

- Implemented a SOTA natural language understanding platform (intent-classification, entity extraction, parsing, & state tracking)- Initiated and oversaw the transition from classical NLP to fine-tuned recurrent neural networks throughout Ambit's product- Built a multi-stage preprocessing pipeline into existing product architecture; decreased preprocessing error rate by 24%- Led research into multi-task, continual learning; enabled us to reuse core feature extraction model for multiple downstream tasks

Auckland, New Zealand

- Co-led the team developing a centralised Django backend for low-latency aggregation of messy, large-scale data sources- Built a network of Django microservices connecting our central server to government infrastructure data sources throughout NZ.- Built an interactive 3D mapping \& visualisation platform for the management of Auckland City Council public health infrastructure.

• Sending a CubeSAT satellite to the International Space Station. Working on the conversion of sensor-data to usable products, database management and the satellite website. • Full-stack Javascript

• Worked with companies like Roomdots, Sodio, Mixmax, & Fashiontap • Business dev activities including market and competitor research, lead compilation, outreach and inbound management.• Investor/ partnership deck creation & campaign design• Financial modelling• Startup hustle

Provide academic and vocational consultation to clients striving for admission to the world's top universities.

• Listened to, analysed and resolved technical issues for customers and clients, including payment processing and Eftpos support, website issues, and contractor demands. • Documented processes and relationships in Microsoft CRM.

Bachelor Of Science - Bs W. Honors, Computer Science, 3.9 Gpa

Activities and Societies: Deep Learning @ Brown research group, Brown Space Engineering software development team, computational.

High School, Dux (Valedictorian) - Highest Ranking Student In Academics.