2. MY CAREER PATH
⢠Sports patterns and translating paintings
to pots
⢠Focused on genetics/neuropsychology in
undergrad
⢠Quantitative MD/PhD student who started
going to geometry/topology workshops
⢠Switched to an MS in Biostatistics after
taking the GRE subject exam
⢠Graduate papers in structural equation
modeling and topology/machine learning
⢠10+ years of consulting and working full-
time in industry data science
⢠Military, healthcare, biotech, education,
marketing, supply chain, consumer goodsâŚ
3. BIG DATA
⢠Volume, variety, veracity,
velocity
⢠Examples:
⢠Wide data that breaks
statistics
⢠Knitting together multiple
sources with different types
of data formats
⢠Non-independent data
⢠Text/image data
5. ANALYZING SURVEYS WITH FLAT DESIGNS
⢠Surveying identity
expression across
social contexts
⢠No hierarchical
structure to allow
factor-analysis-
based solutions
⢠Very small sample
sizes in different
population groups
without evidence of
population similarity
7. ANALYZING SURVEYS
WITH FLAT DESIGNS
⢠Paper linking 0th dimensional
persistent homology with
single-linkage hierarchical
clustering if the distance
metrics are the same
⢠Similar industry problems in
education (testing designs and
validation)
⢠Metrics for persistence diagrams
allow for sample comparisons.
8. PREDICTING EBOLA SPREAD ON TRAVEL
NETWORKS
⢠Democratic Republic of Congo
(DRC) outbreak in North
Kivu/Ituri Provinces
⢠Jungle areas dotted with
settlements
⢠Data included travel routes
connecting settlements and
cases over time in each
settlement
⢠Travel partially do to
populations fleeing rebel attacks
9. PREDICTING EBOLA SPREAD ON TRAVEL
NETWORKS
⢠Graph-based approach to
create a travel network with
case-time information
⢠Opened up Forman-Ricci flow
to highlight changes in travel
and disease outbreak patterns
on the DRC network
⢠Identified areas at most risk
to position supplies (Katwa)
10. TIPPING POPULATION HEALTH BEHAVIOR
⢠Unhealthy behaviors in the
Naval active duty and
reservist populations
⢠Impacts healthcare costs and
deployment fitness of the
force
⢠Large population with local
social ties (units,
neighborhoodsâŚ) and not
many ties over the rest of
the population
⢠Another good candidate for a
11. TIPPING POPULATION HEALTH BEHAVIOR
⢠Identify key vertices
(bridges) in the network for
intervention
targeting/demographics
⢠Estimate impacts via
stochastic differential
equations (SIS model)
⢠Parameters taken from the
extant literature on each
behaviorâs âcatchingâ
properties
12. NUCLEAR CORE LOADING PATTERN
PROBLEM
⢠Reactor core
⢠Rod mixtures
⢠Influences energy field/output
⢠Large optimization search space
(260
)
⢠Qubit approach of quantum
computing/hybrid quantum
algorithms
13. NUCLEAR CORE LOADING PATTERN
PROBLEM
⢠Translation of problem to
quantum computer (D-
Wave)
⢠Ising model
⢠Quadratic Unconstrained
Binary Optimization (QUBO)
mapping
14. NUCLEAR CORE LOADING PATTERN
PROBLEM
⢠Gains from
quantum annealing
over simulated
annealing
⢠Viable solutions
from hybrid
algorithms
⢠Good quantum
annealing
performance
16. CAREER ADVICE
⢠Demonstrated experience:
⢠Publications (match publication history to position needs)
⢠Demonstrate applications to real-world problems (projects, papers,
internshipsâŚ)
⢠Skillsets
⢠Specialize (genomics, NLP, image analytics, video analyticsâŚ)
⢠Programming (Spark/Python/NoSQL lately)
⢠Cloud experience (AWS/Azure⌠through a conference or online class)
⢠Consider a short programming-heavy bootcamp
⢠Indeed or start-up/Angel investing repositories have good
response rates
17. HOW TO CONTACT ME ď
⢠cfarrelly@med.miami.edu
⢠LinkedIn (Colleen M. Farrelly)
⢠Quora (Colleen Farrelly)
⢠Book talks (once the first book is released ď )