Back to portfolio

P29 / engineering project / prototyped

Deep Learning Cycling Posture and Sports Computer Vision

AI, deep learning and computer-vision work for posture and movement analysis in cycling and sports scenarios.

2024-2026UIDE, Ecuador
INTELLIGENT COMPUTATIONAL STACKDeep Learning Cycling Posture and Sports Computer Vision
01DATA SOURCEinstrument / sensor / simulation
02MODEL DESIGNfeatures / representations / data pipelines
03AI / ML / DLsupervised / unsupervised / NLP
04SCIENTIFIC CODEPython / MATLAB / CUDA / GPU
05DEPLOYembedded / full stack / digital twin / XR
06VALIDATEmetrics / experiments / real-case feedback

Engineering data moves through scientific computation, learning algorithms, validation and physical or digital deployment.

01

Problem or Industrial Need

Sports posture and movement assessment can be costly or difficult to scale when it depends on specialized equipment or manual observation.

02

Engineering or Scientific Solution

A computer-vision and deep-learning workflow for analysing cycling posture and movement patterns from visual data.

03

Sebastian's Technical Contribution

Contributed the scientific-computing framing, AI/DL workflow direction and applied engineering supervision connecting model development with a real sports-analysis scenario.

04

Methods and Tools Used

  • Computer-vision pipeline definition
  • Deep-learning model workflow for posture or movement analysis
  • Python-based data processing and model evaluation
  • Applied validation against real-case sports scenarios
05

Prototype, Simulation and Experimental Evidence

prototype

AI posture-analysis workflow

The work is represented as an applied computer-vision and deep-learning project record.

06

Measurable Result or Published Finding

90%

reported cost reduction

The project was positioned to reduce the cost of comparable posture-analysis solutions.

07

Diagrams and Publications

INTELLIGENT COMPUTATIONAL STACKDeep Learning Cycling Posture and Sports Computer Vision
01DATA SOURCEinstrument / sensor / simulation
02MODEL DESIGNfeatures / representations / data pipelines
03AI / ML / DLsupervised / unsupervised / NLP
04SCIENTIFIC CODEPython / MATLAB / CUDA / GPU
05DEPLOYembedded / full stack / digital twin / XR
06VALIDATEmetrics / experiments / real-case feedback

Engineering data moves through scientific computation, learning algorithms, validation and physical or digital deployment.

08

Role, Team Attribution, Institution and Project Context

Applied AI and mechatronics project; scientific-computing and technical-supervision contribution at UIDE.

Connected work

P132024-2026

Multidisciplinary Mechatronics Project Development

Problem
Multidisciplinary student engineering requires a repeatable path from requirements and models to working prototypes and defensible validation.
Solution
A hybrid project-development framework combining conventional engineering control, Agile practices, technical gates and ABET-aligned outcomes.
Evidence / result
35 multidisciplinary projects coordinated
hybrid deliveryABETprototypingtechnical supervision
View system
P142025-2026

NVIDIA Digital Twins, Simulation and Physical AI

Problem
Engineering education and prototype development needed a shared simulation-to-embedded stack for robotics, digital twins and AI-enabled systems.
Solution
Deployment of NVIDIA Omniverse, Isaac Sim and Jetson across simulation, workshops, supervised projects and collaboration as the first official NVIDIA University Ambassador in Ecuador.
Evidence / result
3 NVIDIA platforms deployed
OmniverseIsaac SimJetsondigital twins
View system
P152020-Present

Smart Realities: Spatial Computing and Connected Systems

Problem
Emerging-technology concepts often fail to connect interactive software, physical hardware, learning content and a practical delivery strategy.
Solution
An independent R&D initiative integrating sensors, embedded systems, data processing, spatial interfaces and web-based engineering applications.
Evidence / result
4 cross-functional teams led
XR/ARIoTembedded systemsweb systems
View system