Microfluidic devices are emerging as an attractive technology for
automatically orchestrating complex biology protocols, with
applications ranging from biomedical research to biological
computing. The device technology in microfluidics has been advancing
faster than Moore's Law, with the number of features (valves) per chip
doubling every 4 months. However, current microfluidic chips remain
very inflexible, designed and fabricated for a narrow application
domain.

We are developing new hardware and software abstractions to support
fully programmable, general-purpose microfluidic chips. Our approach
uses a digital design: fluidic samples are discretized into unit
volumes, isolated from one another, and manipulated independently
during operation. Biologists use a high-level programming language
called BioStream to map their own unique experiment to our
general-purpose chips. In this way, a biology experiment becomes a
"program" that can be seamlessly mapped across successive generations
of microfluidic chips.

This talk will provide an overview of microfluidic technologies from a
computer science perspective, highlighting areas in which computer
science researchers can contribute to this field. It will also
describe our recent work in developing new architectures, programming
languages, and CAD tools for the microfluidic domain.