Julia is a programming language created by Jeff Bezanson, Alan Edelman, Stefan Karpinski, and Viral B. Shah in 2009. Since its public release in 2012, Julia has received code contributions from hundreds of scientists, programmers, and engineers across the world.

Today, a small group of researchers reside at MIT and focus on theoretical and numerical aspects of the core Julia language, base library, and several other packages. Current activities center around data science applications, numerical linear algebra, parallel computing, and type theory.

Research

JuliaLabs specializes in collaborating with other groups to solve messy real-world computational problems.

Statistical Genomics

Existing bioinformatics tools aren't performant enough to handle the exabytes of data produced by modern genomics research each year, and general purpose linear algebra libraries are not optimized to take advantage of this data's inherent structure. To address this problem, JuliaLabs is developing specialized algorithms for principal component analysis and statistical fitting that will enable genomics researchers to analyze data at the same rapid pace that it is produced.

This project is an exciting interdisciplinary collaboration with Dr. Stavros Papadopoulos (Senior Research Scientist at Intel Labs) and Prof. Nikolaos Patsopoulos (Assistant Professor at Brigham and Women's Hospital, the Broad Institute and Harvard Medical School).

Financial Fraud Detection

A single stock exchange generates high-frequency trading (HFT) data at a rate of ~2.2 terabytes per month. Automatic identification of suspicious financial transactions in these high-throughput HFT data streams is an active area of research. JuliaLabs contributes to the battle against financial fraud by designing out-of-core analytics for anomaly detection.

Medical Data Analytics

Hospitals, like many large organizations, collect much more data than can be usefully processed and analyzed by human experts using today's available software. Oftentimes, these small-scale analyses can overlook statistical clues that might have rendered substantial improvements to patient care.

In collaboration with Harvard Medical School, JuliaLabs has worked on tools for rapidly identifying potential indicators of irregularities in medical data, equipping doctors and healthcare providers with the analytics they need to make informed medical decisions.

Numerical Linear Algebra and Parallel Computing

JuliaLabs leads the JuliaParallel organization, which maintains the following projects:

JuliaLabs also collaborates with Prof. Steven G. Johnson and Jared Crean in the development of PETSc.jl, a wrapper for the Portable, Extensible Toolkit for Scientific Computation.

Publications

Conferences

People

Current Members

Alumni

Collaborators

The Julia group is grateful for numerous collaborations at MIT and around the world:

Sponsorship

We thank DARPA XDATA, the Intel Science and Technology Center for Big Data, Saudi Aramco, the MIT Institute for Soldier Nanosystems, and NIH BD2K for their generous financial support.

Julia Labs is a member of the bigdata@CSAIL MIT Big Data Initiative and gratefully acknowledges sponsorship from the MIT EECS SuperUROP Program and the MIT UROP Office for our talented undergraduate researchers.