AI software engineer · Malaysia
Yogeswary Mohan
I build AI tools that detect deepfakes, help review code, and turn images into accessible descriptions.

Selected work.
ANIMATED WALKTHROUGH01 / 03
VISTA / VIDEO ANALYSISSIMULATED
SAMPLE VIDEOILLUSTRATIVE FRAME
DETECTION PIPELINE
Extract visual features
Compare across frames
Highlight regions
Regions marked for review
SAMPLE DIFF / REVIEW REQUEST
− return items[0]
+ if not items:
+ return None
+ return items[0]
PARALLEL REVIEW
Logic agent
Edge-case agent
Consensus check
Consensus reached · Approved
SAMPLE IMAGEILLUSTRATIVE INPUT
IMAGE TO NARRATION
Interpret the scene
Apply a narration preference
Generate the description
“A path winds between two hills under a setting sun.”
How I built it
I used a CNN to extract visual features and an LSTM to detect patterns over time. A graph attention network highlights facial regions that contributed to the prediction.
Experience.
-
Mar–Aug 2026
Garmin Malaysia
New-Tech Research Engineer InternAutomated API documentation and worked on camera calibration and storage dispatch for robotic automation.
-
Oct 2024–Jan 2025
Sana’a University
Freelance Software DeveloperBuilt an Android AR app with ARCore, working on real-time 3D rendering and surface detection.
-
2022–2026
Universiti Putra Malaysia
Bachelor of Computer ScienceBuilt Vista, my final-year project on detecting deepfakes in video.