THIAGOMIGUEL
Specialist in Computer Vision and Deep Learning applied to scalable intelligent systems, from research to production.
ENGINEERING
WITH REAL
IMPACT
Computer Engineering graduate from Inatel (2024) with academic distinction, 2nd place in class. Postgraduate student in Machine Learning Engineering at FIAP, focused on MLOps, Big Data and production AI.
I build Artificial Intelligence and Machine Learning pipelines to transform data into intelligent systems in production. I work with computer vision, signal processing and Deep Learning models running on high-performance infrastructure, from raw data to deployment.
APPLIED
SOLUTIONS
TOI: Vital Signs via Video
Complete pipeline for BPM, SpO₂ and RPM extraction via facial video. rPPG + Deep Learning + Anti-Spoofing on NVIDIA DGX H200.
ECG: Morphological Analysis
Advanced digital filtering, R-peak detection and feature extraction. Optuna optimization, ROC-AUC evaluation.
Abdominal Volume 2D
YOLOv8 Transfer Learning for segmentation. 3D geometric modeling from 2D image input.
Datathon: Passos Mágicos
Educational risk prediction model with XGBoost and Optuna. ROC-AUC 0.794 on data from 2,845 students. Full stack with FastAPI, Docker and Prometheus.
TOI — VITAL
SIGNS IN
REAL TIME
System for estimating vital signs from facial video using remote photoplethysmography (rPPG) techniques. The pipeline performs facial detection, dynamic segmentation of regions of interest based on 3D landmarks and temporal extraction of color variations associated with blood flow. Signals are processed by spectral filtering and frequency-domain analysis to estimate BPM, SpO₂ and respiratory rate. Execution runs on NVIDIA DGX H200 high-performance infrastructure with real-time inference.
ECG —
MORPHOLOGICAL
ANALYSIS
Morphological analysis pipeline for ECG signals enabling automatic detection of cardiac patterns. The system performs signal preprocessing, identification of P-QRS-T complexes and extraction of morphological features used by machine learning models. Training is optimized with Optuna and evaluation follows a rigorous protocol with stratified cross-validation and metrics such as ROC-AUC and Kappa.
ABDOMINAL
VOLUME
2D → 3D
Abdominal volumetric estimation system from 2D images using computer vision and geometric modeling. The pipeline employs body segmentation based on YOLOv8 with Transfer Learning, enabling a pre-trained model to be adapted for automatic detection and extraction of the abdominal region. From the extracted mask and body contour, volume is estimated by elliptical and semi-ellipsoidal approximations with refined adjustment via non-linear regression. The approach enables fast, non-invasive estimates with potential application in clinical screening.
DATATHON
PASSOS
MÁGICOS
Predictive model of educational risk to identify, at the start of each annual cycle, students most likely to end the year with INDE below 7.0. The system uses XGBoost with hyperparameter optimization via Optuna, achieving ROC-AUC of 0.794 and F1 of 0.617 on the 2024 test set. Infrastructure includes an API with FastAPI, monitoring with Prometheus and Grafana and three automatic layers of data leakage detection. Project developed for the NGO Passos Mágicos using data from 2,845 students over three annual cycles.
VITAL SIGNS FROM FACIAL VIDEO
ECG SIGNAL
ANALYSIS
Advanced digital filtering + automatic detection of P-QRS-T complexes + extraction of 40+ cardiac morphological features.
ABDOMINAL
VOLUME
2D → 3D
DATATHON
PASSOS
MÁGICOS
PROFESSIONAL
JOURNEY
With a background in electronics and computer architecture, my career evolved towards the application of Artificial Intelligence and Machine Learning in engineering and research projects, combining technical rigor, data analysis and the development of solutions with real-world impact.
Postgrad in Machine Learning Engineering
MLOps, Big Data and AI in production.
Machine Learning Specialist
Computer vision, Deep Learning, HPC/GPU (NVIDIA DGX H200) and AI systems in production.
Academic Mentoring
Digital Electronics I & II, Computer Architecture and Special Topics I.
Computer Engineering
Graduated with academic distinction, 2nd place in class.
Project Analyst
Embedded systems, data acquisition and IoT projects.
Flutter Trainee
Mobile development, UI/UX and backend integration.
FROM DATA
TO DEPLOY
ACQUISITION
Real-time collection via edge sensors. Data structuring and image preprocessing for pipelines.
MODELING
Training Deep Learning architectures focused on high accuracy and low inference latency.
MLOPS
Packaging, orchestration and continuous monitoring. CI/CD for ML models.
UX & PRODUCT
Dashboards that translate complex backends into fluid user experiences.
XAI
LIME/SHAP for prediction transparency. Technical metrics translated into business impact.
TOOLS
& TECHNOLOGIES
- —PyTorch / TensorFlow
- —CNNs · RNNs · BiLSTM
- —Attention Mechanisms
- —GANs
- —Transfer Learning
- —YOLOv8
- —OpenCV
- —MediaPipe Face Mesh
- —Haar Cascade · Dlib
- —Segmentation & ROI
- —FastAPI / Flask
- —Docker · Kubernetes
- —NVIDIA DGX H200
- —MLflow · Optuna
- —CI/CD Pipelines
- —ECG · PPG · rPPG
- —Butterworth · S-G Filter
- —FFT · Band-pass
- —LIME · SHAP
- —ROC-AUC · Cross-val
BEYOND
CODE
TECHNICAL TALK
Presentation on applied Artificial Intelligence, cryptocurrencies, DeFi and the impact of intelligent algorithms on the evolution of financial systems.
HPC/AI INTERVIEW
Interview on NVIDIA DGX H200 infrastructure, GPU architecture, real-time inference pipelines.
It doesn't matter where you start what matters is how you move forward from there.
Thiago Miguel · ML Engineer
LET'S
WORKTOGETHER
Open to ML Engineering opportunities, AI consulting and technical collaborations where data and models drive real impact on product or operations.