Percy Cubas

Percy Cubas

Software Engineer · Robotics, Computer Vision & LLM Systems

GitHub LinkedIn pcubasm1@gmail.com

About

Engineer building full stack platforms for robotics applications, computer vision systems, and LLM integrations with RAG, deployed on AWS. I enjoy working at the intersection of software, robotics, and AI, and I'm looking to bring my experience to new domains.

Python TypeScript C++ React FastAPI Django PyTorch OpenCV YOLO TensorRT ROS-Noetic Jetson Xavier OpenAI API AWS Bedrock LangGraph RAG Qdrant AWS Lambda DynamoDB Docker Linux

Experience

AI Engineer

Oct 2025 - Present
Indra · Peru (remote)
  • Built a web platform to label and correct detections, reducing error on each retraining cycle while maintaining data governance, using ReactJS, FastAPI and AWS.
  • Built a multi-tenant RAG platform with per-session user workspaces and real-time streaming chat for Q&A over documents backed by hybrid search, using ReactJS, Python on AWS Lambda and DynamoDB.
  • Built a document ingestion pipeline for a RAG application, extracting and indexing content from PDFs and scanned documents using DocLayout-YOLO computer vision models, Cohere Embed v4 visual embeddings and OCR.
  • Used CloudWatch logs and metrics to monitor and debug Lambda-based services, tracking invocation errors and latency.
  • Participated in code reviews and wrote unit and integration tests to ensure code quality and maintainability across FastAPI and Lambda services.
  • Migrated an AI-based call-center QA evaluation system MVP to a serverless architecture, using ReactJS, AWS Transcribe (transcription), Bedrock/Claude (scoring) and Lambda.

AI Implementation Engineer

Dec 2024 - Sept 2025
YaVendio · Peru (remote)
  • Implemented AI agents for automated sales with GenAI using LLMs (OpenAI, Claude, Gemini), LangGraph (Python), LangSmith and Prompt Engineering techniques.
  • Automated the integration of product catalogs from Excel files into the AI agents' prompt via a FastAPI service, resolving formatting issues that broke compatibility and replacing manual procedures.
  • Built a CLI toolkit in Python exposed as an MCP server for running internal processes (prompt management, data cleanup, catalog extraction) from AI coding agents (Cursor, Claude Code).

Computer Vision Engineer

Aug 2024 - Aug 2025
Universidad Privada Antenor Orrego · Trujillo, Peru
  • Built the web platform used to manage counting sessions, configure the detection system and run real-time inference, deployed on a Linux services architecture over the Jetson Xavier of an agroindustrial robot, using FastAPI, ReactJS, PyTorch, OpenCV and Unix Sockets.
  • Integrated video streaming to visualize the robot's real-time inference, using WebRTC, ReactJS, TypeScript, FastAPI and Python.
  • Selected the optimal detection model for the robot (agricultural applications) by comparing multiple YOLO versions (v8, v9, v10, v11) through automated cloud experimentation, using Modal and PyTorch.

Full Stack Developer

Mar 2024 - Dec 2024
IDEGO STANDARD · Lima, Peru (remote)
  • Refactored a critical customs order creation service, removing duplicated code and bad practices to improve maintainability, using Python and Django.
  • Implemented frontend debouncing to prevent duplicate requests to the server when a user pressed the same button multiple times, using ReactJS and JavaScript.
  • Automated bank account and transfer validation in Maersk Peru's financial back office, replacing a manual process, using Python and Django.

Research Intern

Sept 2022 - Feb 2024
Universidad Privada Antenor Orrego · Trujillo, Peru
  • Evaluated detection and segmentation strategies for blueberries and artichokes on field images, comparing YOLO and Mask R-CNN.
  • Integrated detection models into a mobile robot, validating the system in agricultural environments, using Jetson Xavier, ROS Noetic and ZED cameras.
  • Deployed an IIoT system to monitor artichoke transplanting, using Linux.
  • Collaborated on the redesign of a PPG optosensor, implementing digital filters to improve robustness against ambient light variations, using C++.
  • Co-author of 5 publications: IROS 2023, 2 at IEEE INTERCON 2023, and 2 at IEEE ANDESCON 2022.

Projects

Robot Platform

Software platform for mobile robots running computer vision applications. Real-time YOLO + BoT-SORT tracking accelerated with TensorRT FP16 on Jetson, live WebRTC streaming, and a web UI to configure counting lines and replay recorded sessions.

Jetson TensorRT WebRTC FastAPI React
Repository

Publications

4 Scopus-indexed (INTERCON x2, ANDESCON x2) + 1 IROS workshop

Deep Learning Approach for Accurate Pre-Harvest Blueberry Ripeness Classification
P. Cubas, R. Huaman, S. Prado
Workshop at IEEE/RSJ IROS, Oct 2023
Artificial Vision Strategy for Ripeness Assessment of Blueberries
P. Cubas, R. Huaman, S. Prado
IEEE INTERCON, Sept 2023
Deep Learning-based Segmentation and Classification System for Artichoke Seedling Grading
P. Cubas, R. Huaman, S. Prado
IEEE INTERCON, Sept 2023
Detection and Classification of Ventura-Blueberries in Five Levels of Ripeness
P. Cubas, E. Fiestas, S. Prado
IEEE ANDESCON, Nov 2022
Paper
IIoT System for Monitoring and Analysis of the Transplanting Process of the Artichoke Seedling
P. Cubas, J. Alva, S. Prado
IEEE ANDESCON, Nov 2022
Paper

Education

B.S. in Electronic Engineering

July 2023 Universidad Privada Antenor Orrego, Trujillo, Peru