Hey There, I'm Elton Roque Lemos

Let me help you get to know me.

Masters Graduate

University at Buffalo

Buffalo, New York.

I'm a Master's of Robotics and AI Graduate from the University at Buffalo. My research interests include robotics software development, Agentic AI, machine learning, computer vision, and natural language processing. I have 3+ years of professional experience as a Product developer in the IT Automations Industry. I have a strong background in programming languages such as Python and C++ with proficiency in frameworks like Django and ROS, and I'm always eager to learn new technologies and techniques.

Top Projects

project thumbnail

Human Segmentation Simulation on NVIDIA Isaac Sim

A full Software-in-the-Loop (SIL) simulation using Isaac Sim and ROS2, where a Nova Carter robot equipped with a camera detects humans in real-time using a U-Net-based segmentation model.

project thumbnail

Synthetic Data Generation using Isaac Sim

A simulation pipeline built in NVIDIA Isaac Sim to generate a synthetic warehouse dataset using domain randomization and train an object detection model using the TAO Toolkit.

project thumbnail

AI Pet Robot

An autonomous pet robot that can recognize, track, follow and interact with its owner.

project thumbnail

F1/₁₀th Deep Racer

Inspired by AWS Deepracer, developed and actor critic modeled car that could compete with a PID controller's lap times and safety.

selected publications

IJASST

Hardware architecture and implementation of an AI pet robot

Elton Lemos, Abhishek Ghoshal, Aditya Aspat

International Journal of Applied Sciences and Smart Technologies 2.2 (2020): 111-136

Summary DOI URL PDF Scholar

The concept of AI companionship is gaining popularity in recent times. Ourproject is an attempt to create a robotic companion that can act just like apet. We created a Pet Robot that can perform various functions like recognisepeople, recognise objects, recognise emotions, play games, follow a personaround, listen to voice commands, dance and so on. This paper focuses on thehardware aspect of the robot and the working of the Core program (Kernel). Itfurther discusses the algorithm that was implemented in this architecture andcompares the final version with other variations. Our work was able to createa fully functioning AI pet robot by using just cheap commercially availabledevelopment boards, motors and a Desktop PC.

IJASST

OpenCV Image Processing for AI Pet Robot

Abhishek Manajit Ghoshal, Aditya Aspat, Elton Lemos

International Journal of Applied Sciences and Smart Technologies 3 (1), 65-82

Summary DOI URL PDF Scholar

The Artificial Intelligence (AI) Pet Robot is a culmination of multiple fields of computer science. This paper showcases the capabilities of our robot. Most of the functionalities stem from image processing made available through OpenCV. The functions of the robot discussed in this paper are face tracking, emotion recognition and a colour-based follow routine. Face tracking allows the robot to keep the face of the user constantly in the frame to allow capturing of facial data. Using this data, emotion recognition achieved an accuracy of 66% on the FER-2013 dataset. The colour-based follow routine enables the robot to follow the user as they walk based on the presence of a specific colour.

IJASST

Stone, Paper, Scissors Mini-Game for AI Pet Robot

Aditya Aspat, Abhishek Manajit Ghoshal, Elton Lemos

International Journal of Applied Sciences and Smart Technologies 3 (1), 65-82

Summary DOI URL PDF Scholar

The Artificial Intelligence (AI) Pet Robot is a combination of various fields of computer science. This paper showcases the various functionalities of our AI Pet. Most of the functionalities showcased use the immage processing modules made available through OpenCV. The pet robot has various features such as emotion recognition, follow routine, mini-game etc. This paper discusses the mini-game aspect of the robot. The game has been developed by using VGG16 convolutional network for identification of the action performed by the user. To improve the accuracy we have made use of background subtraction which gives removes all the unwanted objects from the background and gives a simple cutout of the users hand.