Saswati Ray moving on

[Artur Dubrawski]

Team,

With mixed feelings I need to announce that Saswati will be leaving the Lab
soon. Her last day in the office will be Monday February 7th.

Saswati, what is hard to believe, is a 15-year veteran of the Auton lab.
And (less hard to believe :p ) our most experienced research programmer.
She is one of the few current team members who realizes how much good stuff
exists in the Lab code repository written since 1990-ies in C. After
starting as a system-level programmer, she has been continuously expanding
her skill set to eventually become the queen of multiple Auton things and
projects, including our highly flexible random forest implementation.

Over the past many years she has been managing the development of ML
capabilities in the ERNIE project. In the process, everyone at the prime
contractor - Lawrence Livermore National Laboratory - and the government
sides fell deeply in love with her. Saswati's efforts have helped ERNIE to
beat all the conceivable alternatives so systematically that the system is
on the path to a large scale deployment at every US border crossing now,
and in a recent independent evaluation conducted at about 20 large sea
ports it enabled order-of-magnitude reductions of the need for secondary
inspections of large cargo containers by the US Customs personnel while
significantly improving the probability of detecting real radiation threat
vs the pre-existing systems.

More recently, Saswati led engineering efforts to develop our automated
machine learning tool called Auto^nML. Her common-sense-driven approach to
conditioning data for analysis and simple but effective model selection
procedures helped make Auto^nML a persistent winner of multiple DARPA
competitions over the past few years. I used to equate that accomplishment
to winning 7 or more Super Bowls or Stanley Cups in a row. Indeed, it is
not a small feat to beat so systematically many highly capable competitors
from lead AI universities, even though Saswati and our DARPA D3M/MILEI team
have been equally systematically sharing their code with the community.
Yet, Auto^nML was still victorious at the next event.

Saswati has many more impressive accomplishments under her belt, and she is
undeniably one of the nicest colleagues I had a pleasure to work with.

So, it is with mixed feelings indeed, sadness prevailing, that I write this
note.

I am sure we will all wish her all the best with her next endeavor, and let
her know that in case she ever gets bored or not challenged enough, she
should know her way back.
Once an Autonian, always an Autonian.

Cheers,
Artur

PS We will be having a small but memorable farewell party for Saswati;
time, place and format TBD.

PSPS Some of you will shortly be asked to engage in capturing Saswati's
workload so that we will have no disruptions in our deliverables and
other project obligations.
And, Everyone, if you please can think of anything that we should do to
brace for impact before Feb 7, please let me and Saswati know.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/autonlab-users/attachments/20220125/105e43aa/attachment.html>