Dr
.
Elena
Voss
Artificial
Intelligence
Engineer
·
Research
Scientist
elena
.
voss
@
alumni
.
stanford
.
edu
+1 (650) 555-0192
San
Francisco
,
CA
linkedin
.
com
/
in
/
elenavoss
github
.
com
/
elenavoss
AI
Engineer
with
8+
years
of
experience
designing
and
deploying
large
-
scale
deep
learning
systems
at
Google
and
two
venture
-
backed
startups
.
Specialise
in
NLP
,
representation
learning
,
and
production
ML
pipelines
.
Published
12
peer
-
reviewed
papers
(
NeurIPS
,
ICML
,
ACL
)
with
2,400+
citations
.
Passionate
about
building
models
that
work
reliably
in
the
real
world
—
not
just
on
leaderboards
.
EXPERIENCE
Senior
AI
Engineer
Jan
2023 —
Present
Cortex
AI
·
San
Francisco
,
CA
ML
Research
Engineer
Jun
2020 —
Dec
2022
Google
Research
·
Mountain
View
,
CA
Led
design
and
production
deployment
of
a
1.3
B
-
parameter
transformer
-
based
document
understanding
system
,
reducing
document
processing
time
by
82%
across
6
enterprise
customers
.
—
Built
a
distributed
training
pipeline
on
64
A
100
GPUs
using
PyTorch
FSDP
and
DeepSpeed
,
cutting
training
time
from
21
days
to
4.5
days
.
—
Developed
an
active
-
learning
feedback
loop
that
improved
F
1
by
14
points
over
three
deployment
cycles
with
fewer
than
5,000
labeled
examples
.
—
Mentored
a
team
of
4
ML
engineers
;
established
code
-
review
and
experiment
-
tracking
standards
that
reduced
regression
rate
by
60%.
—
Designed
a
sparse
mixture
-
of
-
experts
architecture
for
multilingual
NLU
that
achieved
state
-
of
-
the
-
art
results
on
14
of
18
XTREME
tasks
while
reducing
FLOPs
by
37%.
—
Created
and
open
-
sourced
a
200
k
-
sentence
synthetic
data
pipeline
for
low
-
resource
language
pairs
,
adopted
by
3
external
research
groups
.
—
Deployed
a
real
-
time
toxicity
detection
model
serving
500
M
+
queries
per
day
across
Google
Products
,
improving
precision
from
0.81
to
0.93
at
constant
recall
.
—
Authored
5
NeurIPS
publications
and
2
patents
related
to
efficient
attention
mechanisms
and
cross
-
lingual
transfer
.
—
1 / 3
Machine
Learning
Engineer
Aug
2017 —
May
2020
Lumina
Analytics
(
acquired
by
Datadog
) ·
New
York
,
NY
EDUCATION
Ph
.
D
.
in
Computer
Science
2013 — 2017
Stanford
University
·
GPA
: 4.0
Dissertation
: "
Efficient
Representation
Learning
for
Structured
and
Multilingual
Text
."
Advisor
:
Prof
.
Christopher
Manning
.
Stanford
Graduate
Fellowship
.
B
.
S
.
in
Computer
Science
&
Mathematics
2009 — 2013
University
of
California
,
Berkeley
·
GPA
: 3.94
Highest
Honors
·
Phi
Beta
Kappa
·
Regents
'
Scholar
.
SKILLS
ML
Frameworks
PyTorch
,
TensorFlow
,
JAX
,
Hugging
Face
Transformers
,
Ray
,
DeepSpeed
,
ONNX
Runtime
Languages
Python
,
C
++,
CUDA
,
Rust
,
TypeScript
,
SQL
Infrastructure
Kubernetes
,
Docker
,
Terraform
,
MLflow
,
Weights
&
Biases
,
Vertex
AI
,
SageMaker
,
Airflow
Domains
Natural
Language
Processing
,
Representation
Learning
,
Large
Language
Models
,
Anomaly
Detection
,
Multi
-
modal
Learning
,
Efficient
Deep
Learning
Research
Experimental
design
,
literature
surveying
,
paper
writing
,
peer
review
(
NeurIPS
,
ICML
,
ACL
program
committee
)
PROJECTS
Adaptive
Token
Merging
for
Long
-
Context
Transformers
PyTorch
·
Triton
·
FlashAttention
Developed
a
learned
token
-
merging
module
that
reduces
KV
-
cache
memory
by
40%
with
less
than
0.5%
accuracy
drop
on
LongBench
,
enabling
128
k
-
token
inference
on
a
single
A
100.
Built
the
core
anomaly
-
detection
engine
for
a
real
-
time
observability
platform
,
processing
12
TB
of
telemetry
data
per
hour
with
sub
-
second
latency
at
p
99.
—
Implemented
a
variational
autoencoder
for
multivariate
time
-
series
that
detected
34%
more
incidents
than
the
threshold
-
based
system
while
reducing
false
-
positive
rate
by
4×.
—
Designed
a
feature
store
in
Redis
and
Bigtable
that
cut
model
training
iteration
time
from
hours
to
minutes
.
—
2 / 3
Multilingual
Legal
Document
Summarization
Transformers
·
LoRA
·
XLM
-
R
Built
a
zero
-
shot
summarization
system
covering
14
languages
for
a
pro
-
bono
partnership
with
the
Stanford
Legal
Design
Lab
,
achieving
ROUGE
-
L
scores
within
3
points
of
English
-
only
performance
.
PUBLICATIONS
SimBa
:
A
Simple
Baseline
for
Multilingual
Sentence
Embeddings
NeurIPS
2025 ·
Voss
,
E
.,
Chen
,
T
., &
Liang
,
P
.
Mixture
-
of
-
Experts
for
Cross
-
Lingual
Transfer
:
A
Pareto
Analysis
ICML
2024 ·
Voss
,
E
. &
Kudugunta
,
S
.
Efficient
Fine
-
Tuning
via
Structured
Pruning
of
Attention
Heads
ACL
2023 (
Outstanding
Paper
Award
) ·
Voss
,
E
.,
Sanh
,
V
., &
Rush
,
A
.
Anomaly
Detection
in
High
-
Dimensional
Telemetry
with
Variational
Recurrent
Models
SIGKDD
2021 ·
Voss
,
E
.,
Park
,
J
., &
Gupta
,
A
.
+8
additional
peer
-
reviewed
papers
(
NeurIPS
,
EMNLP
,
ICLR
) –
full
list
at
elenavoss
.
com
/
publications
2,400+
total
citations
·
h
-
index
: 14
HONORS
Google
Ph
.
D
.
Fellowship
in
Machine
Learning
(2015–2017) ·
NeurIPS
Outstanding
Reviewer
Award
(2023) ·
NSF
Graduate
Research
Fellowship
Honorable
Mention
(2014) ·
Best
Paper
Award
,
Stanford
AI
Lab
Symposium
(2016).
SERVICE
Program
committee
member
:
NeurIPS
(2022–2025),
ICML
(2023–
2025),
ACL
(2023–2025),
ICLR
(2024–2025).
Reviewer
for
JMLR
,
TACL
,
and
IEEE
TPAMI
.
Co
-
organizer
of
the
Workshop
on
Efficient
Natural
Language
Processing
at
NeurIPS
2024.
Mentor
for
Women
in
Machine
Learning
(
WiML
)
mentorship
program
since
2021.
3 / 3
