Kayky Matos

G

Global Scalability

Jul 2024 - Current (1 year 2 months)

My role included scaling our workflow and tools to match the new heights that the company was reaching, replacing limited tools for data and machine learning with Big Data tools for our ETL process to send to the model like Spark on EMR and updating to use tools like AWS Sagemaker and Amazon Lex

C

Cost Optimization of the ChatBot and the Model

Jul 2024 - Current (1 year 2 months)

My role included updating and monitoring the code of the Model and the Chatbot to optimize costs by reducing the number of processes and tokens used in a task

I

The Future of Technology and Data: The Fusion of Generative AI and Quantum Computing

IDCN-AI

Jun 2024 - Jul 2024 (2 months)

C

Continuous Learning & Application

Jun 2024 - Current (1 year 3 months)

by implementing AI solutions and integrating AWS services like Glue and App Studio, conducting live lessons on both for the teams leading to streamlined data workflows. I stay current with emerging AI technologies and apply new knowledge to improve platform functionality. I evaluate and select cost-effective AI models and services, balancing performance with cost-efficiency. I lead initiatives to optimize data processing resources, supporting the platform's global growth.

I

The Future of Technology and Data: The Fusion of Generative AI and Quantum Computing

IDCN-AI

May 2024 - Jul 2024 (3 months)

Modeling Concrete Compressive Strength with Deep Learning and Neural Networks using Keras

IBM

Jan 2023 - Dec 2024 (2 years)

M

Mitigating Bias with Model Remediation in TensorFlow

Jan 2023 - Jan 2025 (2 years 1 month)

A

AWS Lambda with Amazon CloudWatch and SNS to Implement a Slack Chat Bot

Jan 2023 - Nov 2024 (1 year 11 months)

In this project, I developed a Slack chat bot from scratch that utilizes AWS Lambda, Amazon CloudWatch, and Amazon SNS to deliver real-time notifications to a Slack channel whenever specific CloudWatch alarms are triggered. The project focused on a text classifier for identifying toxic content, employing machine learning and the MinDiff method to mitigate bias against vulnerable groups. Technologies used: Python; NumPy • pandas ◦ TensorFlow • Keras The project aims to create a document search engine using Multimodal Retrieval Augmented Generation (RAG) to improve the retrieval of textual and image information. The system extracts metadata, performs contextual searches, and generates integrated responses, facilitating navigation in compl

M

Machine Learning Model for Predictive Analysis of Financial Markets

Jan 2023 - Sep 2024 (1 year 9 months)

moving averages. Additionally, I assessed investment risk by examining daily return distributions and inferring average returns from several years of historical data. Technologies: Python, SciPy, Matplotlip, Pandas, Numpy, statistical models. Developed this project to explore the financial transactions data for fraud analysis, apply feature engineering and machine learning techniques to detect fraudulent activities using BigQuery ML. Technologies: GCP, BigQuery ML, SQL I developed a concrete compressive strength prediction model. Implemented and optimized a deep learning regression model for model evaluation and train/test split. Achieved robust performance through extensive testing and hyperparameter tuning, demonstrating expertise in mach

Image Captioning with Visual Attention Using Neural Networks

Google

Jan 2023 - Nov 2024 (1 year 11 months)

The project focuses on developing an image captioning model that generates textual descriptions from input images using a neural network architecture based on an encoder-decoder framework with visual attention. Initially, the COCO dataset is prepared and processed with TensorFlow and its associated libraries, ensuring images are resized and normalized. A pre-trained InceptionResNetV2 model is employed as a feature extractor in the encoder, which generates embeddings that are passed to the decoder for caption generation. The attention mechanism enables the model to concentrate on various parts of the image during this process. Tokenization and preprocessing of the captions are also critical steps, involving the addition of special start and

Building an Interpretable TensorFlow Image Classifier with Vertex AI

Google

Jan 2023 - Dec 2024 (2 years)

E

Emotion Detection Project with Python and Flask

Jan 2023 - Aug 2024 (1 year 8 months)

Building a RAG system for web data using Llama 3.1-405b and LangChain

IBM

Jan 2023 - Oct 2024 (1 year 10 months)

The goal was to develop an application that would generate real-time contextual responses by extracting information from a web-based knowledge repository. The initial phase involved setting up the development environment on watsonx.ai. The second phase involved installing libraries such as LangChain and IBM Watson Machine Learning. The third phase focused on indexing URLs to build a knowledge base. Content was extracted and cleaned using LangChain’s UnstructuredURLLoader, with documents organized for efficient processing. Text was split into manageable chunks using RecursiveCharacterTextSplitter, optimizing for a chunk size of 512 characters. The fourth phase involved generating vector embeddings with the IBM Slate model, which were

Developing a Chat Application Using LLM and RAG with AlloyDB and LangChain

Google

Jan 2023 - Oct 2024 (1 year 10 months)

This project aimed to create a chat application leveraging Retrieval Augmented Generation (RAG) and Large Language Models (LLM). The first phase involved setting up the development environment on Google Cloud and AlloyDB, a PostgreSQL database optimized for AI workloads. I activated Cloud Shell, installed the PostgreSQL client on a virtual machine (VM), and connected to the database. In the second phase, I created and initialized the database called "assistant", enabling the pgVector extension for efficient manipulation of vector data. The third phase focused on installing Python and its dependencies, setting up a virtual environment for application development. The subsequent phase involved populating the database with relevant dataset

M

Machine Learning Fraud Detection on Financial Transactions

Jan 2023 - Sep 2024 (1 year 9 months)

M

Machine Learning Regression Model using Keras

Jan 2023 - Sep 2024 (1 year 9 months)

Sentiment Analysis Model with BERT and Deep Learning

Google

Jan 2023 - Oct 2024 (1 year 10 months)

M

Multimodal Retrieval Augmented Generation (RAG) using the Vertex AI Gemini API

Jan 2023 - Oct 2024 (1 year 10 months)

Deep Learning Image Classifier Using ResNet18 and PyTorch

IBM

Jan 2023 - Sep 2024 (1 year 9 months)

P

Python Projects

Jan 2019 - Jan 2025 (6 years 1 month)

About

Detail

Timeline

Résumé