ETL, Data Mining, Data Analysis, Business Intelligence
Resume:
Marek Gajdosik
ETL/BI DEVELOPER/SOLUTION ARCHITECT/PRE-SALES CONSULTANT
Pelhrimovska 1194, Dolny Kubin 026 01, Slovakia
Phone : +421-***-***-***
e-mail : *****.********@*****.***
Summary
I am a highly experienced consultant/data analyst with focus on business
intelligence. Strong business analysis, revenue assurance, fraud & risk
management and account management skills, as well as a strong IT
background. Extensive experience in providing complex and critical software
solutions to many enterprise customers. A proven track record in R&D,
solutions design, solutions implementations, data management and
professional services. Extensive telecommunications knowledge.
Specialties: data analytics, revenue assurance, fraud detection, risk
management, financial solutions, project management, oracle applications,
data Extraction, Transform and Load (ETL) solutions, migration projects,
Lavastorm solutions and other development projects, Guru in tools like
Pentaho/Informatica/Alteryx.
Skills
Business and Soft Skills
. Self-motivated, highly organized, creative and detailed oriented
. Knowledgeable - within Telco industry especially, Insightful - high
motivated to learn and experience
. Communicative and helpful team-player, autonomous if required
. Management, Professional Services, Coordination and Management with Key
Stakeholders
Technical and Analytical Skills
. Data Analysis, Data mining, Big Data, Data Transformation, ETL, Data
Modelling, Data Warehousing, Data Integration, Reporting, Dashboards
. Business Intelligence, Agile Methodologies, Business Analysis, Solution
Architect, Algorithms, QA Testing, Operations
. Telecommunications, Revenue Assurance, Fraud and Risk Management,
Mediation, Billing, Quality Assurance (QA), Controls and Risk, Trading
. Unix, Microsoft Office, Microsoft Excel, Microsoft Word, JasperSoft,
iReport, other Microsoft Producst related to Analytics like SSAS, SSRS
etc
. SQL, Oracle, Hadoop, PL/SQL, MySQL, MSSQL, R, NoSQL, T-SQL, PostgreSQL
. PHP, Java, Python, XML, jQuery, Shell, Regular Expressions, Xpath, CSS,
HTML5
. Lavastorm Analytic Engine (LAE) Expert - Data Analytics Tool (tool
similar to Pentaho, Alteryx, Informatica etc.), Lavastorm Resolution
Center (LRC), Lavastorm Transaction Warehouse (LTW)
Intermediate knowledge only
Work Experience
LAVASTORM ANALYTICS INC.
PRINCIPAL CONSULTANT (04/2009 - 11/2014)
. Designed, developed and delivered 20+ complex Revenue Assurance and
Fraud & Risk management solutions to many enterprise customers
. Support with Solutions Architecture, RFPs, POCs for Pre-Sales team
. Author of Best Practices Guide and co-author of Training Guide for LAE
. Trained 80+ customers' analysts in Lavastorm Analytic Engine (BI Tool)
. Handled 10+ installations and innumerable upgrades of LAE
. Incorporated two newcomers providing guidance on their first client
facing project
. Became "one of the best, if not the best LAE resource" / "the best LAE
much better than others" within 3-4 years based on my former colleagues
opinion
. Majority (70-80%) of the projects listed below I implemented myself, at
client site, attending key stakeholders meeting as the only Lavastorm
consultant
. Identified 450 000,- EUR monthly revenue loss due to systems
inconsistency for KPN
. Avoided 7 mil. loss by reconciling replacement mediation system with
current one for Ziggo
Projects (for full list see the linkedin profile):
KPN Netherlands: Arrow
Project Arrow
Position Consultant, Designer, Developer
Client KPN Netherlands
From - to 1/2013 - 11/2014
Project Using log files the CDR counts are verified daily. For some
descriptio sources the log files were not possible, so the raw files had
n to be used. This is monitored at different levels starting from
switch until end of billing. The verification checks % of
filtering within mediation system, billing filtering and
rating, day over day changes and other elements. This
application is meant to find out any issues within entire
lifecycle of CDRs. Thresholds set are the key for alerting.
Technology SQL, Oracle, Unix, Shell, JasperSoft iReport, Lavastorm
Analytic Engine (ETL Tool), Lavastorm Transaction Warehouse
(LTW)
Responsibi Initial Consultation with client on approach and solution
lity architecture. Once agreed on my proposed solution I offered
only ad-hoc help to the other consultants who implemented this
solution. Creation of generic log and data reader, which was
then adjusted by other team members specifically for each
stream. After deployment I fixed also few remaining JasperSoft
Dashboard defects, which others could not deal with. Automated
script written in shell to allow repeated execution of the
solution was created by me as well.
Ziggo Netherlands: Mediation and Rating Verification
Project Mediation and Rating Verification
Position Consultant, Designer, Developer
Client Ziggo Netherlands
From - to 2011 - 2014
Project Ziggo first wanted to put in place KPIs around their current
descriptio mediation and rating system. So the shadow mediation and shadow
n rating control was built for the residential market. The check
was daily on high level, just checking the trending and
thresholds, but if there was any inconsistency, full detail
shadow mediation and rating was run to identify the issue
exactly. Later on during the migration from old mediation
system to new one (Comverse->Comptel) and during migration from
old billing platform to new, the migration process was
controlled by the very same solution, which was adjusted to
account for current needs only. During this project by
identifying incorrect filtering rules in mediation system the
deployment of new mediation system was postponed and roughly 7
mil. EUR loss was avoided. Estimate is based on not billing
valid CDRs due to filtering rules in mediation system till the
end of the year.
Technology SQL, Oracle, Python, Unix, Shell, Lavastorm Analytic Engine
(ETL Tool), Lavastorm Transaction Warehouse (LTW), XML
Responsibi Entire end to end solution was designed and created by me only,
lity from gathering requirements, development, deployment, testing
etc. Only MS Excel report layout for daily/monthly reports was
designed by someone else. First extraction of data using python
from files with different formats depending on the source
origin was required. After extraction, data was cleansed and
analysed based on the mediation/rating rules. Even though the
specs were solid, still further analysis of results and data
mining was required and based on the outcome, further iterative
adoption of the rules was implemented to achieve 100% accuracy.
Especially within rating, plenty of results did not follow any
rules provided. Shell script which allowed automated or manual
run of the entire solution was created, allowing to run the
control with different parameters.
Kabel Deutschland, Vodafone Ghana: Least Cost Routing
Project Least Cost Routing
Position Consultant, Designer, Developer
Client Kabel Deutschland, Vodafone Ghana
From - to 2012 - 2012
Project Interconnect providers have different prices for different
descriptio locations. Telco providers want to use the cheapest settings.
n Based on the input pricelists the best settings for routers
needs to be determined. LCR solution is provided for very high
prices and takes lengthy time to implement at client site.
After roughly 3 months of development on my own, I created
a solution which was able to process 9 different pricelists
from different providers (9 out of 9 required) and using very
simple configuration that could be managed by any Excel user,
the LCR configuration was determined within few moments. Entire
implementation would require less than a week of single
consultant on client side. Leaders in providing LCR claim to do
this within couple of weeks/months, with team of consultants
onsite. Additionally for Vodafone the comparison of ICT prices
with retail prices was done identifying certain destinations,
where company was actually paying more than invoicing their
customers.
Technology SQL, Oracle, MS Excel, Lavastorm Analytic Engine (ETL Tool),
Lavastorm Transaction Warehouse (LTW)
Responsibi Entire logic and all algorithms on dividing the ranges and area
lity codes were created by me. Additional features after consulting
with SME were added by me again (exception list, quality
exclusions, fixed settings etc). Data was provided in
differently formatted pricelists from Interconnect providers,
thus extraction, cleansing and manipulation mainly to allow
identical area format was required. Python was heavily used in
combination with data mining and very deep analysis in further
step where area codes were converted to ranges instead,
allowing to price each range with multiple prices. Due to
involvement of LTW and its own area codes the solution got even
trickier and SQL language querying Oracle LTW tables had to be
involved. After determination of cheapest provider for each
range again the backwards conversion from ranges to area codes
was required to configure switches properly.
KPN Netherlands: Customer Data Retention
Project Customer Data Retention
Position Designer, Developer, Tester, Solution Architect, Consultant
Client KPN Netherlands
From - to 5/2011 - 6/2014
Project Based on the Dutch law, Telecom providers were obliged to allow
descriptio government and state offices to query their customer base for
n last 6 months. KPN being the largest Telco provider in
Netherlands with plenty of legacy systems being still alive had
a huge challenge, since all of these systems had to deliver
response based on the query sent, i.e. from police.
Technology Python, SQL, Oracle, Unix, Shell, Lavastorm Analytic Engine
(ETL Tool), Lavastorm Transaction Warehouse (LTW), Regular
Expression, Hadoop, Hive, Sqoop
Responsibi More than 25 systems delivered customer data on daily basis as
lity flat files. Since the data could not be manipulated in any form
during loading process, many checks for quality for each field
had to be implemented. Depending on the data issue, either
entire file or specific records were rejected and reported.
Only valid records complying exactly with specs were loaded and
used for querying part. To ensure only valid data regular
expression was used heavily (IPv4, IPv6, emails etc.). Shell
and Cron was used to regularly check for incoming request
files. Based on this request all Oracle tables loaded
previously are queried returning corresponding results. Some
queries consisted of sub-queries where two or three
tables/systems were queried based on the results of previous
one. This very complex solution was designed and implemented by
me. Other team members contributed on loading part of the data
into the Oracle and created reporting on data quality based on
rejected records. The complexity of queries was especially in
the design of data. Some source systems were daily snapshots
while other provided Create/Delete event records. Moreover
single customer consisted of multiple records within each
system, one for name, one for address and one or many for each
subscription within system. The complex correlation of these
records based on datetimes and IDs was additional challenge.
Moreover the validity of the records could not be handled in
loading part due to requirements, so I had to deal with invalid
scenarios, like two consecutive creations without delete or
similar. Due to security, the encryption with GPG of the
requests and answers had to be implemented as well. This was
handled with python which was used also within LAE to start
execution of the query only if request file was valid. Later on
due to unexpectedly high volumes of data it was decided to
migrate the solution from Oracle to Hadoop. To keep the effort
as low as possible Sqoop and further Hive was used to maintain
the original SQL statements usable. Here was my first
experience with Hadoop infrastructure for roughly 6 months.
This was mainly migration process of the current solution.
Kabel Deutschland: Interconnect Shadow Rating
Project Interconnect Shadow Rating
Position Consultant, Designer and Developer
Client Kabel Deutschland
From - to 5/2010 - 5/2011
Project KDG had their interconnect rating provided by 3rd party. Still
descriptio they wanted to check if rates applied by this Swiss provider
n were accurate and correct.
Technology SQL, Oracle, Unix, Shell, Lavastorm Analytic Engine (ETL Tool),
Lavastorm Transaction Warehouse (LTW)
Responsibi Reimplementation of complex rating rules was my task, and being
lity on my own on the project, I had to create reports as well.
These reports consist of detailed records which were identified
as incorrectly rated and as well as aggregated reports on
overall validity.
Since the specification on rating was poor or none sometimes,
reverse engineering and heavy data mining was the only way to
identify the rules based on the data provided. Using reference
data and CDRs with rates applied I had to analyse and identify
very complex rules, such as trunk ownership, timeband and
pricelist determination and then destination/price assignment.
After implementing and documenting all these, KDG provided this
document for confirmation of validity to ICT provider. All the
reference data was pulled using Oracle DB link/SQL, while CDRs
were read in as flat files. CDRs required certain cleansing in
order to be able to identify destination based on number
prefix. Entire solution was run daily using shell script I
created.
Telstra Australia: Indirect Channels
Project Indirect Channels
Position Data Analyst
Client Telstra Australia
From - to 2010 - 2010
Project Telstra as provider sells their service through 3rd parties
descriptio using provisioning system. Due to false sales and fraud, the
n analysis of the data from this segment was required, to provide
baseline for modifying the rules within this market.
Additionally Telstra was reconsidering certain channels and
provisions for these and analysis of the past data was the key
for their decisions and adjustments.
Technology Python, Oracle, Lavastorm Analytic Engine
Responsibi I took over this project after data extraction and cleansing. I
lity was cooperating with Telstra business expert, where he
explained and outlined different issues/targets and my
responsibility was to find out answers within data. The aim was
to make provisions for different channels consistent. Due to
outer effects, the provision got unequal throughout the time.
Additionally various resellers behaved fraudulent by making
fake agreements, which were cancelled afterwards etc. My task
was not only to answer questions asked, but additionally to
identify by data mining any additional issues or adjustments
that could improve this market segment. I provided raw data at
the end for Telstra expert to use them within Excel reports.
Windstream, USA: USOC Code Verification
Project USOC Code Verification
Position Consultant, Designer, Developer
Client Windstream, USA
From - to 1/2010 - 3/2010
Project Billing using USOC code is very complex and specific USOC code
descriptio can be applied only if the equipment fulfills specific
n criteria. Based on complex rules, invoices were verified, if
applied codes are correctly assigned. Daily 5+ people were
manually checking invoices for validity going through dozens of
invoices. The solution allowed to have this check done
automatically and to configure the rules by the users
themselves.
Technology Python, Unix, Shell, Java, Lavastorm Analytic Engine (ETL
Tool), JSP
Responsibi Implementation of rating rules and afterwards reporting.
lity Reporting was implemented in web-browser GUI (Java/JSP), where
configuration of the billing rules was possible to certain
level as well as fixing the incorrect values using workflow.
Data analysis and data mining was the key to identify rules
applicable. Afterwards these were consulted with Billing
experts. Shell script handled automated daily runs.
O2 United Kingdom: SimBox detection
Project SimBox detection
Position Consultant, Designer and Developer, Data Analyst
Client O2 United Kingdom
From - to 2011 - 2011
Project SimBoxing is illegal and fraudulent behavior in United Kingdom.
descriptio The task was to identify such usage and report it, based on
n pure CDRs extract.
Technology Python, Unix, Shell, Lavastorm Analytic Engine (ETL Tool)
Responsibi Based on the Call Data Records provided by O2, data analysis
lity was required to identify this type of fraud. After discussing
SimBoxing with SME within our company, I came up with scoring
solution. First data was extracted using python. Data cleansing
and manipulation was done to let through only fields required
to check simboxing, to minimize the volume. There are various
behaviours identified during simboxing. For each of these, the
subscriber was checked and for each behavior he was scored
achieving final score at the end. Allowing thresholds
configuration for each rule, in the output report were only the
most probable fraudsters. Default configuration had to be
determined by data analysis and further results confirmation.
KPN Netherlands: IBC Epacity
Project IBC Epacity
Position Consultant, Designer, Developer
Client KPN Netherlands
From - to 2011 - 2011
Project The Installed Base Control for Epacity product range is to
descriptio provide ongoing reporting of the alignment between the service
n instance and associated products installed in Cramer, Portal,
and Siebel. These three systems represent the technical,
billing, and CRM systems respectively.
Technology SQL, Oracle, Python, Unix, Shell, MS Excel, Lavastorm Analytic
Engine (ETL Tool)
Responsibi Creation of reports that will include discrepancies in the
lity service status and installed products. The details will be
analyzed to investigate the root causes of the discrepancies
and the counts and monthly amounts will provide insight into
their impact. I took over the reconciliation development
somewhere in the middle and had to finish the graph along with
defining the discrepancy rules with Key Stakeholders. After
data extraction, cleansing and manipulation was unavoidable
prior joining different systems, since level of records
differed (physical/logical) across systems.
KPN Netherlands: End to End Install Base Reconciliation
Project End to End Install Base Reconciliation
Position Designer, Consultant, Developer
Client KPN Netherlands
From - to 5/2009 - 5/2010
Project KPN asked for reconciliation of their Install base for DSL
descriptio products. The entire reconciliation involved 27 systems, which
n needed to be joined together to provide insight into the
customers subscriptions.
Technology Java, SQL, Python, Unix, Shell, Oracle, Lavastorm Analytic
Engine (ETL Tool), JSP
Responsibi Design of logic for merging all the systems together (Technical
lity (3x), Order Management (2x), Billing Systems (5x) plus
additional like complaints, migration etc.). Three way
comparison was used. The quality of the data after extraction
was very bad and thus cleansing was required in order to allow
joining of the sources. These had to be enriched with Customer
complaints systems, Migration track keeping systems and many
others to provide the reports which included the details on
inconsistencies in between these. The overall monthly loss
identified due to inconsistencies was ca. 450 000,- EUR. The
reporting tool was designed and implemented by me as well.
Java/JSP was used to create these reports along with Oracle DB.
Alltel, USA: Reports processing
Project Reports processing
Position Consultant, Designer, Developer
Client Alltel, USA
From - to 5/2008 - 6/2008
Project Processing of reports without any standard formatting and
descriptio loading these into DB and displaying in front-end GUI. Multiple
n different items are covered, such as errors from different
sources, prepaid, postpaid audits, SMS audits, switch to bill
audits etc.
Technology SQL, Oracle, Python, Unix, Shell, Java, Lavastorm Analytic
Engine (ETL Tool), JSP
Responsibi Python was used to extract, parse and cleanse the data.
lity Creation of generic reader for 27 different reports varying
only very slightly was at the beginning. Making this reader
configurable allowed to read in all of them using this single
reader. Data quality was checked further as the reports
contained both details and summaries (quality was achieved by
comparing these). Based on requirements data had to be
manipulated (rotation, aggregation) to get desired view. At the
end data was loaded into Oracle. Web-browser GUI (Java/JSP/SQL)
allowed to view these reports.
IBM GLOBAL SERVICES
BUSINESS INFORMATION ANALYST (02/2009 - 04/2009)
. Automated manually generated reports
. Pushed through AutoIt language to be used as automation technology
CONNEX-N TECHNOLOGIES
IT CONSULTANT (12/2006 - 11/2008)
. Implemented Revenue Assurance Solutions for telco providers, i.e.
France Telecom
References
DAVID WROBEL
COO at Cyfeon Solutions
FORMER VICE PRESIDENT OF PROFESSIONAL SERVICES at LAVASTORM ANALYTICS
LIESBETH PHILIPPI
MANAGER PROGRAM DELIVERY AT LIBERTY GLOBAL
FORMER DIRECTOR OF PROFESSIONAL SERVICES EMEA AT LAVASTORM ANALYTICS
NEIL DUFFIELD
PROGRAMMA DIRECTOR AT TIGO TANZANIA
FORMER PROJECT MANAGER AT LAVASTORM ANALYTICS
THOMAS TRULSSON
Settlement Expert at Telia Denmark
+45 261 001 64
Contact this candidate