resume
callum dempsey leach, m.sc. · lead data and software engineer · github.com/caldempsey · [email protected]
engineering leader who has helped raise over a hundred and fifty million dollars in funding across ai cybersecurity, social media, climate science, healthcare and music royalties. a decade of building data-intensive distributed systems for clients across emea and north america: from founding a startup acquired by a nasdaq-listed corporation, to leading mlops infrastructure at one of the most decorated ai companies in the united states.
experience
- principal engineer and tech lead (mlops) · reality defender, new york · february 2024 to now
- defending businesses and governments from manipulated media. a 0-to-1 data lake at 100 mb/s and 100m+ rows/s ingestion for model training and evaluation; a strongly typed streaming rest api for olap with schema evolution; java databricks toolchains reverse-engineered into go at 1,575× the official drivers’ throughput; ci/cd, a three-tier secured network, soc2, hipaa and gdpr auditing; lineage, licence and source-poisoning tracing; c-suite advisory. golang, databricks, spark connect, apache arrow, terraform, aws, delta lake, unity catalog, kubernetes.
- senior software engineer (contract) · stage.tech (Buma Stemra), united kingdom · november 2023 to may 2024
- brought in to rescue a stalled royalty distribution platform under contract with buma stemra, the netherlands’ collecting society for composers and publishers (40,000+ members, founded 1913, distributing €251m+ a year). brought on to motivate the team to be better as much as to build: a project two years in went from 13% complete to 55% in a month; an scd type 2 audit system tracking billions of daily royalty records. java, aws, lambda, snowflake, postgresql.
- lead platform engineer (contract) · mucoaid (Berlin Institute of Health), berlin · january to august 2023
- an ml platform improving physicians’ oral cancer detection by 80%. legacy typescript microservices refactored to 95% behaviour-driven coverage; the devops initiative moving everything from gcp to a secure aws architecture on terraform. golang, terraform, cassandra, kubernetes, nats, just, rds.
- platform engineer, mlops (contract) · unweave.io, berlin · june to july 2023
- multi-node, multi-gpu training on cloud tenants: four key features in a month, among them sftp transfer 75% faster through compression, repository synchronisation and the build configurations for k8s-managed vms, shared volumes and multi-gpu instances. golang, terraform, aws, postgresql, the kubernetes and docker apis.
- lead platform engineer · geneva (Bumble), new york · june 2019 to november 2022
- acquired by bumble inc. (nasdaq: bmbl) in 2024 for the strength of its technology platform. a social messaging app from conception to 100k+ daily active users and a $12m series a; engineering scaled from 7 to 50+ people; ten-plus event-driven features; the go and data function: a worker system at 60,000 jobs/s, a ring buffer caching 500 mb of database objects per instance, 10±5 ms responses, redis pub-sub load-tested to 3m concurrent users, a data lake lifting retention 30%. golang, typescript, dynamodb, elasticache, ecs, s3, sns, sqs, terraform, aurora, grpc, kinesis, datadog, athena.
- data engineer · the oakland group (Vodafone), leeds · february to june 2019
- vodafone’s enterprise-wide migration off legacy teradata warehouses onto redshift: 300m rows a day with 100% migration assurance, in a two-week pilot. scala, python, teradata, databricks, emr, redshift, glue, spark.
- software engineer · rhea group, oxford · february 2018 to january 2019
- an early warning weather system for uganda’s farmers under a £127m european space agency programme: go pipelines for 114 gb/month of geospatial data, python etl for 4d image pyramids, an elixir web-socket api streaming map visualisations at 25 ms from sources at 4 gb/s. golang, elixir, kafka, nifi, kubernetes, postgresql.
- technical support consultant · emis health, leeds · december 2015 to september 2016
- the patient records of the nhs: t-sql deployment configuration spanning 90% of the clinical systems in england and wales.
- data centre sre tool (contract) · dxc (Hewlett Packard Enterprise), washington · december 2013
- a python tool configuring 200+ network switches across the estate, saving eighteen days of network engineers’ time.
public works
apache spark · apache arrow · databricks · golang-migrate · elixir · gocraft/work · the rest.
education
master of science, computer science · newcastle university · 2016 to 2017 · colin graham award (accenture-sponsored), best group project in software engineering, as the team’s technical lead.
bachelor of arts (hons), philosophy · university of leeds · 2012 to 2015.