Engineer at @Lacework. Formerly Microsoft Azure, Google Brain, MSR Silicon Valley and University of Cambridge.
Redwood City, CA
Joined February 2007
- Tweets2.3K
- Following397
- Followers3.5K
- Likes1.1K
Derek Murray retweeted
HotOS begins tomorrow (11AM EST)! We have 30 wonderful papers and 4 panels. Check out the program, papers, and talk videos: sigops.org/s/conferences/hot…
Derek Murray retweeted
Have proven experience in DL *and* software optimization? My team at Google Brain has Research SWE openings. Looking for folks interested in making compilers, databases, networks, and kernels smarter. RL/Bandits experience a plus. DM me and/or RT.
Derek Murray retweeted
The HotOS program is out. We have an awesome set of papers! Please check it out and register!
@ACMSIGOPS HotOS's program is available now: sigops.org/s/conferences/hot…. Don't miss out on these great works! Register at: cvent.me/0K4XW9. Registration is free for students and affordable otherwise. Thanks to @bariskasikci, @xexd, and the PC for setting up a great program!
Derek Murray retweeted
Are you interested in AI runtimes, AI kernels, using GPUs to accelerate AI workloads, and looking for a job with a great and friendly culture?
Come join me and our team!
These jobs are US-remote friendly, ignore the "Redmond" tag for now:
careers.microsoft.com/us/en/…
Derek Murray retweeted
From 2017-2018, I worked on TensorFlow 2 as a full-time engineer on the Google Brain team.
My impressions: debugmind.com/2020/01/04/pat…
(Happy to answer questions about my experience, or about deciding to leave Google for a PhD)
Derek Murray retweeted
New blog post about research opportunities for #ML data storage & preprocessing, including lessons from a study of real ML workloads at Google, in collaboration with @jsimsa and @mrry. More details in our paper: arxiv.org/pdf/2101.12127.pdf
What challenges does modern ML pose to storage systems? Scalabilty, yes, but also ordering and freshness guarantees. sigarch.org/rethinking-data-…
Derek Murray retweeted
AI friends working on TensorFlow and PyTorch: if my team could build you *one* debugging tool, either standalone, or integrated into VSCode or Jupyter what would it be?
Derek Murray retweeted
A new advanced guide on optimizing input pipeline performance using the TF Profiler and tf.data (thanks @rachelim!)
About the TF Profiler: blog.tensorflow.org/2020/04/…
New performance guide: tensorflow.org/guide/data_pe…
Derek Murray retweeted
Prof. Malte Schwarzkopf (@ms705) talks about the origin of dataflow computing, how it has been applied to big data problems, and how it could be used to tackle new challenges such as GDPR compliance!
sigops.org/2020/the-remarkab…
Derek Murray retweeted
New meeting - Scott Sanderson talk on — @frankmcsherry COST and Differential Dataflow. Monday October 7, Microsoft New England Research and Development Center, in Kendall Square.
meetup.com/Papers-We-Love-Bo…
Derek Murray retweeted
Presented tf.data and tf.distribute at #GoogleMLSummit in Tokyo! Stay tuned for a recording.
Derek Murray retweeted
🎉 Excellent! Thanks so much for the contribution, Yong Tang! 😁
This release will help folks who are using tf.contrib.ignite/kafka/etc. to plan their transitions to @TensorFlow 2.0.
@apachekafka
👋 Have opinions?
Want to participate?
Join a SIG here! tensorflow.org/community/lis… – at Los Angeles, CA
This Post is from an account that no longer exists. Learn more
Derek Murray retweeted
Our #VLDB18 OtterTune demo pits the world's best DB researchers against machine learning. The competitors have to tune @postgresql's knobs for TPC-C. So far the #TensorFlow algorithms are winning.
Demo paper from @bohanzhang14 + @danavanaken: vldb.org/pvldb/vol11/p1910-z… #VLDB2018
Derek Murray retweeted
Today we released and open sourced Pulumi, a new way to build modern cloud software using real languages. Read more on my blog joeduffyblog.com/2018/06/18/… or download it at pulumi.io.
Derek Murray retweeted
In @TensorFlow 1.9, it is much easier to use Keras with the Data API: just pass data iterators, specify the number of steps per epoch, and you're good to go! Plus it works in both graph mode and eager mode, kudos to the TF team!
Derek Murray retweeted
The original TensorFlow control flow ops (the ones underlying `cond` and `while_loop`) are detailed in a fun new paper: arxiv.org/abs/1805.01772
But they were likely a mistake: if/for/while can be lowered to functional control flow instead, which is easier to implement+parallelize
Derek Murray retweeted
TPUv3! So hot it needs help with cooling: first liquid cooled devices in our data centers..
A TPUv3 pod is 8X as powerful as the TPUv2 pod announced at IO last year, offering more than 100 petaflops of ML compute, allowing us to tackle bigger problems & build better products. xitter.cf/Google/status/99390386…
Today we're announcing our third generation of TPUs. Our latest liquid-cooled TPU Pod is more than 8X more powerful than last year's, delivering more than 100 petaflops of ML hardware acceleration. #io18
Derek Murray retweeted
My talk (Training Performance) from the @TensorFlow #TFDevSummit is live at: youtu.be/SxOsJPaxHME You can download the Cloud TPU traces and analyze them yourself in TensorBoard from goo.gl/5a1zMH
