Connect models to inputs
Camera streams, barcode scans and hand landmarks become inputs to a complete application, with practical constraints visible in the case studies.
COMPUTER ENGINEERING BACKGROUND / TARGET ROLE
AI / ML Engineer
Models become useful when they meet a real input, a clear decision and a working interface.
Computer Engineering background at UTeM with applied project work in object detection, hand-gesture classification and multi-model orchestration. Connects model inputs and evaluation to practical interfaces, data storage and observable system behaviour.
HOW I COULD CONTRIBUTE
From the degree
to the work.
Read evaluation honestly
Precision, recall and cross-validation are presented with their test context, alongside occlusion, lighting and user-environment limits.
Make behaviour inspectable
Consensus routing, telemetry and recovery mechanisms make the agent pipeline easier to review when providers disagree or fail.
SELECTED / AI / ML ENGINEER
The work behind the fit.
Three projects selected for this role. Open a case study for implementation details, constraints and evidence.
Computer vision · Capstone
Hybrid Self-Checkout
- Built a Python pipeline combining YOLOv8/OpenCV product detection, USB barcode input and MySQL product/transaction data.
- Repository-reported evaluation: 77.4% precision and 72.0% recall after 50 epochs; tested visual verification challenges including occlusion and similar packaging.
77.4% precisionReported model benchmark · 50 epochs
Explore the case study ↗
Human interaction · Browser ML
Gesture recognition
- Converted MediaPipe hand landmarks into 71-dimensional features and a 100-tree Random Forest for browser inference in JavaScript.
- Implemented gesture navigation with cooldowns and virtual controls. Reported 98.6% cross-validation accuracy; not a real-world user accuracy measurement.
98.6% cross-validationReported score · not real-world accuracy
Explore the case study ↗
Agent systems · Edge infrastructure
Hermes
- Built a documented Python/Termux experiment with multi-model consensus routing, telemetry, watchdog recovery and post-mortem analysis on mobile hardware.
- Separated serving and gateway processes and used telemetry to review model disagreement and unavailable-provider behaviour.
Consensus + recoveryExperimental system · contribution submitted
Explore the case study ↗RELEVANT TOOLKIT
Skills, in context.
Applied ML
YOLOv8, OpenCV, MediaPipe Hands, Random Forest, feature engineering, precision/recall, cross-validation.
Programming & integration
Python, JavaScript, SQL/MySQL, camera and barcode inputs, browser inference.
Agent systems
Multi-model consensus, LLM orchestration, Termux/Linux, telemetry, watchdog recovery.
PROFESSIONAL REGISTRATIONS
Engineering, with credentials.
BEM Graduate Engineer
Computer branch · Approved 15 October 2026.
Verify in the BEM public directory ↗MBOT Graduate Technologist
Information & Communication Technology (IT)
GT26100872 · Approved 22 October 2026.
Explore another direction.
Each view starts from the same Computer Engineering background and highlights different project evidence.
NEXT / A CONVERSATION
Have a role
in mind?
I’m looking for a team where I can contribute, learn and build useful technology.