My job as Machine Learning Engineer and Tech Lead was guiding the team to design and implement in production a Machine Learning model using Python and Cloudera tools to automate an EUC. Click the button for details.
Projects' DescriptionDuring my Master's, I had the opportunity to collaborate in two different lectures. In the first one, I created assignments regarding the collection and analysis of data in real-time, using Kafka and Dockers, to train ML and DL models (Online Learning). In the second lecture, I was in charge of helping and guiding students through the design and implementation of a User Interface using HTML/CSS JavaScript (Bootstrap and React).
Projects' Description
In this area, we analyzed the client's needs and suggested the optimal solution for them to automate their processes using bots. Our bots were programmed with Automation Anywhere, and they lived within Virtual Machines.
Projects' DescriptionDuring my first rotation in the "Citi Bank Technology Leadership Program," I collaborated with the Architecture department. We were in charge of verifying that the technologies and frameworks used in all projects were correctly structured and their versions were safe for the Bank. In this area, I was in charge of validating Design Charters and automating processes using Visual Basic Analysis (VBA).
Projects' Description
This was an internship and my first job in a domotics enterprise. My responsibilities were designing the audio-visual solution together with an architect. Afterwards, I started to program the devices to automate lights, audio and video. The devices used were Crestron and Lutron.
The purpose of this project was to process batch information from the client stored in a HUE database and, using a pre-trained Gradient Boosting model, determine whether that specific client should have a payroll loan. Technologies used are Python, Cloudera, HUE as the database, Spark, and Autosys as the scheduler. We also used Agile methodology, reporting all steps in JIRA and storing the code in a private GIT.
My role in this project was first a developer, translating code from SAS to Python and Spark, but then I became tech lead; I was in charge of administrating together with the PM the activities of the developers and having meetings with other areas to ensure the correct deployment of the product to production servers.
The purpose of this project was to use supervised learning in batch to process once a month information from 11 million clients and store the result obtained from a Gradient Boosting algorithm, saving the data in Hue Database and then into Mongo DB to be consumed by other areas.
For this project, my team and I created a series of Data Quality Rules (DQRs) to ensure the data was clean and ready to process. Afterwards, I was chosen to be the L3 support engineer for the complete pipeline, which was the most challenging part.
This project aimed to reduce the time branch officers invested in checking information of specific clients when they requested a loan. Here, we used Automation Anywhere to create bots that, based on data from the customer, performed a lookup in different systems and databases to obtain all information and leave it handy for the officer.
My duties here were to program in Automation Anywhere several bots to look for the information in different systems. But the most challenging part was asking for permission from the owner of the system to allow the bots to enter and look for information as if it was a user. So, my team and I needed to ensure the bot complied with all the bank's politics.
This area was in charge of validating that all technological projects use the optimal combination of technologies and systems. My job here was to record each project and help project managers deal with deadlines. I programmed a series of automated emails whose trigger was based on a platform path.
This project was my first baby. It started with a POC and me acting as a Business Analyst, understanding its needs. The project involved analyzing Excel files and passing them to an HDFS platform in Hadoop. Once there, they were used to feed a Random Forest every six months to create a model capable of predicting which clients who were initially rejected for getting a personal loan should be reconsidered for a smaller loan. This project was done using Python and Cloudera.
My role in this project was as a Business Analyst to understand our client's needs. Then, as a programmer, I prepare the data for extraction, variable selection, feature engineering and model training. My final role was as the Tech Lead to load the solution into a productive server.
The purpose of this project was to use a Machine Learning model to classify different geographical locations within Mexico City based on the number of businesses that are around to better plan opening or closing a branch office or ATM locations. For this project, we used Autosys, Python, and Cloudera tools such as HUE and Tableau.
This project was my second baby. When I got there, the model was already created, but I had to optimize the code, as it took more than 7 hours to run. Also, ensure the ingested data meets all the security standards for being stored in a safe database. Also, I was in charge as Tech Lead to organize the correct communication with Tableau for showing results and preparing the productive server to store raw and processed data together with the algorithm safely.
This project decreased the risk of an EUC (a manual process). It gathered information from different Excel files and concatenated them into a single file. Then stored it inside a secure database to be consumed by the users. My role in this project was to automate using Automation Anywhere the merging process. After the project was done, it was my responsibility to train over 300 users all around the country. So, I did a series of video tutorials with examples and the most relevant information on interacting with the bot.
The area gave three levels of support to critic applications inside the bank. I was a Level 2 support engineer, and my job was to monitor alerts in a platform and try to solve the problem ASAP. Otherwise, it should be escalated with the Level 3 support engineer.