Data Scientist
Total years of experience :10 years, 3 Months
Innogy is a energy renewable company that set up
an innovation hub to develop artificial
intelligence projects
1. Deep learning LSTM recurrent neural networks applied to
predictive maintenance of wind farms
2. Credit score rating system for an energy company( work in
progress)
Technologies used: Python(scikit-learn, pandas,
tensor flow), R, SQL, spark.
Breakthrough is a media agency that has a lot of their
content in social media platforms. I was hire to build
a daily data gathering system of that content from
different social media platforms.
1. Google app engine application built using python that
captures data using different API connections and stores it on
a MySQL database and NoSql.
2. Web application developed in django to classify sentiments
on tweets and Facebook comments with the aim of building
training sets for machine learning models.
3. Natural language process model built in python and spark to
analysis semantics on tweets and Facebook comments.
4. Natural language process model to perform sentiment analysis
utilizing neural networks.
5. D3.JS dashboard with PPC performance indicators and traffics
estimations
Technologies used: Python(scikit-learn, pandas),
R, SQL, Spark
I was hired by Hitachi to collaborate on the
development of a proof of concept that involves
building a prioritisation system for containers
that arrive at a port.
1. Supervised learning approach making use of python sci-kit
learn to train and test support vector machines
2. Semi-supervised learning approach with support vector
machines developed in R
3. D3.js dashboards to visualise models recommendations
Technologies used: Python(scikit-learn, pandas),
R, SQL, SQL, D3.JS
My duties at Microsoft were to build an automated
system that can classify and analyse both text and
images on a gamers feedback platform. I also
involved on Microsoft Paint telemetry system.
1. C# Natural Language processing max entropy approach to
approve or reject user's feedback regarding its content.
2. Developing and training a support vector machine
to perform sentiment analysis on user's
feedbacks. This task was accomplish with python
pandas and scikit-learn
3. Usage of Speeded up Robust Features algorithm and
a support vector machine to detect and classify
bugs during gameplay on screen shots.
4. Usage of map and reduce jobs, SQL queries to extract and
transform data on Azure to later be presented in a PowerBi
dashboard.
Technologies used: C#, Java, R, Python, Excel and SQL,
Hive.
I was hired by IG to devise and implement attribution
models across multiple channels. I was responsible for
the mathematical modelling and implementation of this
project. As IG Data Scientist my main responsibilities
where:
1. The first phase revolved around collecting
cookie level noisy data and creating a SQL clean
database with clusters.
2. The second phase focused primarily on using machine
learning, Bayesian network and Markov chain methods implemented
in Java to
model the channels and their dimensions
3. The final phase used Markov decision processes and genetic
algorithms to processes and genetic algorithms to optimize the
channel spend to maximize the profit of the advertising
campaign. This was accomplish using JAVA.
4. The next step was to build a simple Hadoop cluster to
implement and train the machine learning models making use of
map-reduce jobs.
Technologies used: Java, R, Python,
Excel, Hadoop, Map reduce, Pig.
Simply Business is an online business insurance
broker.
I was involved in projects within different teams
of company always with the aim of automating and
optimizing processes.
1.
Taki
ng
advant
age R to
PP
C
key
wor
d
of
perfo
rm s
cl
us
te
ri
n
u
s
i
n
g
K -
means
and
clus
teri
ng
s
p
a
t
i
a
g l
to
mar
ket
opportu
nities
unde
rst
ide
nti
fy and
a
n
d
which
ones
where
performing
b
a
d
l
y
Hubskip - Hubskip was an airfare predictions
startup that the main aim was to create a product
similar to a financial option on a flight ticket.
As the company only Data Scientist I was
responsible for all the mathematical modelling.
1
.
Develo
p calibrations
optio
ns
mode
ls
f
o
r
in
proc
esse
s
an
d
bas
ed
Levy
stochastic
differential
equations.
2
Bu
il
d
regr
essi
on
mo
de
ls
t
o
pr
ed
ic
t
th
e
.
airline fares
3. Volatility predictions using time series models and
stochastic process such GARCH and HESTON model.
Technologies used: Python, R, Matlab,
Excel and Cassandra.
a e
s
m
yins
deve
ltop
applications to allow the communication between
the cars and the databases.
1. Build Python scripts to send packets from the car to a
mongoDB using UDP and TCP/IP protocols
2. Infer and interpolate GPS coordinates using a small
sample file and greedy algorithms
3. Develop a web application to receive packages from the
car and plot it on a map with the engine features.
4. Build java classes to manipulate files using FTP protocol
5. Build a simple python API to interact with a OBD can
adapter.
Technologies used: JavaScript, Java, Python, CSS,
HTML.