Marcos Gabriel Guimarães De Assis Email & Phone Number

Rio De Janeiro, Rj, Br

Rio De Janeiro, Rio De Janeiro, Brasil

Rio De Janeiro, Brasil

- Organization of the company's data in a serverless data centralization platform using AWS stack (Athena, Glue, S3, and Redshift). The modeling was responsible for maintaining historical data of managed assets (heavy machinery), facilitating preparation for maintenance events, and of the tools used in maintenance, reducing the monthly cost of acquiring new ones by 5%;- Creation of a Telegram bot to help prepare technical reports on machine conditioning, reducing production time by 3 hours. The report is created automatically as the maintainer is guided through expected activities for each type of maintenance using a checklist;- Predictive analysis of tooling used for each type of maintenance and heavy machine based on maintenance history, which reduced the time taken to choose the tools by 1 hour;- Creation of an IoT system to collect and process data from electrical positioning sensors and electrohydraulic valves for remote analysis of assets managed via AWS Cloud. An MQTT communication between an electronic circuit with ESP32 and the AWS IoT Core service is established, processing data in real-time via the AWS Kinesis service, storing it in an AWS DynamoDB database, and visualizing it via an AWS QuickSight Dashboard. This pipeline has allowed the company to identify faults more quickly, cutting costs by 25% with unnecessary trips to the site to repair something small.

Fortaleza, Ceará, Brasil

- Implementation of a unified and scalable storage infrastructure, allowing data integration from various sources and easy access by employees from different areas. A serverless data lake architecture implemented on the AWS cloud platform was used, ensuring a solid foundation for data analysis and other digital services for the company;-Development of a photovoltaic system monitoring platform to facilitate remote fault identification. Data from the inverter's sensors and weather stations was collected in real-time through processing pipelines via the Databricks platform and used to train machine learning models (SARIMAX and LSTM) to identify faults at different times (day, month, and year). This resulted in an 11% reduction in corrective maintenance costs;- Automation of repetitive tasks using Selenium and Botcity, such as generating performance reports for photovoltaic systems, commercial proposals, and daily performance reports for employees. As well as automating the process of analyzing a client's credit against a catalog of partner banks. This allowed employees to focus on more strategic activities, reducing operating costs by an average of 32%;- Descriptive analysis of the company's different marketing channels using SQL via AWS Athena service. Data from Google Ads, social networks (Linkedin and Instagram), and email was analyzed, to compare financial and customer acquisition costs for each channel with their returns on some performance indicators, such as conversion rate, sales volume, average ticket, and click-through rate. According to the marketing team, this helped to optimize the campaign's budget allocation.

Fortaleza, Ceará - Brasil.

- Responsible for the implementation and documentation of instrumentation tests of photovoltaic modules, such as open circuit voltage measurement, short circuit current measurement and measurement of characteristic curves of photovoltaic modules;- Responsible for the implementation and documentation of tests to identify anomalies in photovoltaic modules, such as thermography and electroluminescence image processing techniques;- Responsible for monitoring and assisting the maintenance of the 5 photovoltaic systems of the IFCE - Fortaleza campus;- Determination of KPI's to quantify the quality of energy production in the system;- Creation of a dashboard with Plotly / Dash to increase the flexibility of control over assets, allowing the automatic identification of more cloudy and sunny periods (days, months and years);- Creation of a WebApp connected to an REST API for the diagnosis of photovoltaic modules, using thermographic and electroluminescent images together with neural networks (Mask-RCNN and CNN) and digital image processing techniques;- Application of regression models (ARIMA and LSTM) in time series data from the inverter's outputs of each system in periods of 6 months, which guaranteed the purchase of 2 more photovoltaic systems by the institution.