Hi! I'm David Tench, a computer science researcher interested in processing massive datasets.
I design space-efficient, memory-hierarchy-aware algorithms and build systems to address massive-scale problems in areas like bioinformatics, databases, network measurement, and machine learning. I am particularly interested in designing and implementing practical graph streaming and sketching algorithms.
Recent News
Here's what I've been up to lately.
GraphZeppelin followup: Check out my paper on Landscape, a distributed graph processing system which uses linear sketching techniques to overcome the communication costs that typically bottleneck distributed systems. To appear at ALENEX 2025. Check out our code here.
Did you know that Bloom (and quotient, cuckoo, etc) filters are vulnerable to malicious attacks? My upcoming SIGMOD 2025 paper introduces adaptive quotient filters , which fix their mistakes to defend against adversaries and improve performance on naturally-occurring database and networking applications; essentially for free!
I was recently awarded the 2023 Grace Hopper postdoctoral fellowship at Lawrence Berkeley Labs , funding a two-year postdoctoral position to build systems powered by randomized and memory-hierarchy-aware algorithms to solve massive-scale scientific problems. Read my interview!
About Me
I am the 2023 Grace Hopper postdoctoral researcher at Lawrence Berkeley Labs advised by Aydin Buluç. Before that, I was an NSF Computing Innovation Fellow at Rutgers University advised by Martin Farach-Colton and Michael Bender. I completed my PhD at UMass Amherst in the College of Information and Computer Science, where I was advised by Andrew McGregor. I design algorithms and build systems for large-scale computation, in particular graph sketching algorithms for graphs that are massive, dense, and change over time. I apply these algorithms and systems to real-world problems in areas like databases, bioinformatics, network measurement, and machine learning.
Contact
Email me at dtench [at] pm [dot] me.