Daniel Roca

I created the reporting solution for the healthcare company. I began with the data architecture and data modeling of the reporting solution. Afterward, I made the report visualization and relevant KPI related to employee performance, revenues, and referrals.

I was in charge of creating a set of reports and visuals that would help the marketing team assess the effectiveness of their Google and Facebook advertising campaigns, including all the funnel metrics that would help them be more effective in their financial services. The solution was created using Looker and Snowflake.

I created a data pipeline to extract hospitality data of a Company, and then I modeled the data to assess the price of their rooms at a given time. The output of the data pipeline where some dashboards that recommended the price that the company should ask for their properties based on the historical reservations.

The project involved creating multiple Power BI reports identifying the credit insurance behavior of clients, including their premium value and associated risk. The ETL processes used for the dataset were also created using Power BI dataflows.

My contributions to this project were the Power BI reports and adapting the ETL architecture to Power BI dataflows.

The aim was to create an end-to-end tool that extracted football results from an API and transformed and stored them to create a data warehouse using Pentaho.

This data warehouse was used as a decision tool to create descriptive dashboards using Power BI and machine learning models that aimed to predict the results of the matches of the main European leagues.

I contributed to this project by defining the architecture, creating the database, implementing the data warehouse, and creating the ETL processes, reports, and machine learning models.

A whole pipeline of data transformations from Excel files to create a data warehouse with relevant taxing information from many companies in the United States.

After combining the different data sources, we created the data model, reports, and dashboards with Power BI. Finally, embedded Power BI was used to create a web page with these embedded reports that will serve the clients' needs to understand their taxation behavior and design future strategies or amendments.

My contributions to this project were some ETL processes using Microsoft Azure Data Factory and the dashboards with minor Power BI administration tasks.

Creation of 30-hour lessons with demos to help beginners understand the benefits of using data and analytical techniques that solve various business needs.

These lessons also included a data analytics learning path with Microsoft tools, including Microsoft Azure, SQL Server, and Power BI certifications.

To better understand client behavior patterns and offer them better advertising, the clients were segmented using a lifetime value (LTV) approach and then clustering techniques using k-Means and DBScan.

With the created clusters, each of these groups received focused advertisements intended to be more relevant for them and increase the app's usage.

The project began with all the information of clients that had not used the app anymore. Based on their features, we created an ML model that gave them a score of being active users again after some incentives.

This ML model was used to choose the clients with higher probability and focus the incentives on them to increase the app revenues.

The project scope was to select the best advertisements for the clients. To choose suitable clients for each ad, we developed SQL stored procedures that ran daily. These stored procedures chose the clients based on the variables such as location, age, and previous purchases to identify the best advertisement and the best time to send it.

The project had two phases. The first one was to extract the data from multiple sources to find all the locations where renewable solar and eolic energy projects could be located.

The second phase consisted of using a data visualization tool to find the best locations based on the characteristics that were extracted.

The project consisted of migrating the data from legacy systems to SAP. The scope was to ensure that source data was migrated correctly. For this task, we used Rapid Miner to validate the data quality rules expected by SAP tables and create the spreadsheets used in SAP.

The project's scope was to extract named entities (Enterprises, People, and Locations) from government contracts and create a graph structure to identify recurrent relations in contracts.

Since these contracts are given to the best technical offers, the aim was to identify anomalous behavior.

The approach was to create a Named Entity Recognition using Azure Cognitive Services and to create an interface that displayed the graph of the most recurrent relationships in contracts.

Diverse machine learning models were created to predict the delays of infrastructure projects in the Colombian government. The approach was to use the historical data in a classification model to predict in the future which infrastructure projects would be delayed based on their descriptors.

The machine learning models were created using Python and the sci-kit-learn library.

The project consisted of updating a government entity's data architecture to find tax evaders. They had legacy systems that were only used to store the data and not to use it for the decision-making process.

We proposed different architecture and business intelligence tools to leverage the data they had and the one they would handle afterward.

The project defined the structure and protocols for sharing and publishing data among government entities. The main goal was that the inhabitants could access each entity's relevant data to make revisions.

My contributions to the project were to define the data architecture of the sharing protocols and the database design to define how the data would be stored.

A data warehouse that extracts financial data from over 800 entities of the Colombian government, models their financial KPI and calculates the level of debt of each one. After consolidating the data warehouse, we created reports that enabled analyzing the data within different dimensions, such as period, region, and source ministry.

My contributions to this project were the ETL processes for the debt data mart. Also, the related indicators and the Power BI reports for this topic.

The financial execution of a company was reported using Microsoft Excel. The project consisted of automating the data extraction from the source files to transform them into a standard tabular form that would enable further analysis with a dashboard tool.

VBA Macros were made to develop the data transformations, and Power BI was used to create the dashboards to review the most relevant financial KPI.

The project was to create the cost hierarchy for a healthcare company and a way to better account for costs. This cost structure was based on leading practices for healthcare. The scope of the project was:
• Diagnose the current state.
• Proposed information architecture model.
• Data Governance model.
• Roadmap.

The project consisted of reviewing a Microsoft Project file representing the SAP implementation for healthcare. The schedule had more than 15,000 activities and more than 80 resources. The main activity was automating the responsibility of delays by using VBA and presenting the information with a data visualization tool.

The project aimed to create different regression techniques from scratch, using Python to allow users to benefit from other techniques and compare the results these techniques gave them depending on the data. The regression techniques implemented were vanilla, lasso, ridge, and robust regression.

Finally, the project included an implementation of a decision tree from scratch used for regression and classification. This implementation's add-on could also be used to plot the decision tree.

By using public streaming data and transforming it into a dashboard that could be used to understand the best hours, locations, categories, and content, we achieved a higher probability of publishing videos that became trending videos on the streaming platform.

My contributions to this project were transforming the source files using Python and implementing the reports.

Using deep learning techniques for image recognition to identify workers' locations in construction sites enabled finding the riskiest places and implementing security measures.

The image recognition framework was based on the state-of-the-art model detectron2 implemented initially by Google.

Use of deep learning techniques and specifically a U-Net architecture to segment objects from satellite images. These permitted us to assess the damages from a hurricane that impacted several towns in the United States of America.

Prepared a teaching sample for a business analytics course that the client plans to launch fully. Since this was just a teaching sample, it is not a complete project. The teaching sample consisted of data integration and cleaning.

I developed an app that refreshed automatically every day and gave the users the daily tracking of productivity of employees, referrals, staff retention, and turnover, among other relevant KPIs.

I automated the data extraction from SharePoint to Power Query and then to Power BI, and I also automated the scheduled refreshes of the reporting app.

Finally, I managed the user authentication and licensing to enable access only to a portion of users.

This is a 12-hour e-learning course that includes a project, the course content, videos, exercises, and quizzes. The topics included in this course were machine learning (classification, regression, and clustering) with Power BI, time series forecasting with Power BI, and AutoML and Power BI dataflows.

I integrated a Power BI solution with different APIs and Python code, updating the dashboards with internal data and external learning data and creating a full perspective of the company's position internally and externally at a team and company level.