The TensorFlow team at Google invited me to give a talk on Remora's programming model, type system, and open problems. A preliminary visit earlier in the week had brought up data frame computation, so this is the first exploration of adding heterogeneous records to Remora as a way to work with data frames. While the thesis proposal is currently the most thorough overview of the Remora project, the REPL-session style used in much of this talk is probably easier to understand without the speaker.
Workshop talk for the ARRAY 2018 paper
Part 2 of a 2-part series based on the ESOP 2014 talk, describing the design of Remora to the team while interning at Nvidia.
Part 1 of a 2-part series based on the ESOP 2014 talk, describing the design of Remora to the team while interning at Nvidia.
Conference talk for the ESOP 2014 paper
This was an early "project status report" while visiting INRIA Rocquencourt during my advisor's sabbatical there. Since I had a fairly long time slot to fill, this includes a lot of worked-out examples of evaluating rank-polymorphic code as well as a tour of the formal reduction rules and some typing rules.