Diffix, Jan'19–present, Security & Privacy, Paul FrancisInvestigated compatibility of commercially deployed private data analytics system to ML algorithms. Tested the system’s resilience to specific types of privacy attacks.UniTraX, Apr’16–Dec’18, Security & Privacy, Deepak Garg, Paul Francis, Matteo MaffeiDeveloped methods to improve yield on limited query budgets without reducing result accuracy. Designed and implemented budget tracking solely based on queries, which allows to publish tracking state without the danger of leaking sensitive information.Towards Low Noise and Large Budgets, May’15–Dec’18, Security & Privacy, Paul FrancisDesigned systems that provide graceful deviation from privacy guarantees in order to allow more queries with less noise on results.Giving Up on Privacy Guarantees, Jun’14–Dec’18, Security & Privacy, Paul FrancisInvestigated practical alternatives to “Differential Privacy”. Worked towards support of unbounded numbers of queries with bounded result noise.Triage, May’13–May’14, Distributed, dependable, and mobile systems, Allen Clement, Björn BrandenburgImproved user satisfaction through more accurate backend response time prediction. Selective execution of requests with good prognosis raises system goodput.Multi-KV, Nov’12–Aug’13, Distributed, dependable, and mobile systems, Allen ClementBuilt mechanisms for consistency in geo-replicated storage systems.Work Stealing at Scale, Jan’12–Oct’12, Programming languages and systems, Umut AcarScaled work stealing schedulers to clusters of machines.

Designed data layer for serverless computing platform and evaluated performance of existing distributed storage systems in the Autonomous Software Systems Research department under supervision of Volker Hilt.

Worked on key-value store efficiency and phase change memory adaption in the FAWN systems group under supervision of David Andersen.