Linh
Tuan Ung

Teaching AI to evaluate speech mechanics in real-time.

• Engineered WebSocket streaming for sub-second pronunciation scoring.
• Deployed and optimized edge ML models to classify phonemic variations.
• Secured and hosted the public API strictly within a single edge GPU constraint.

Moving data across complex schemas without losing a single packet.

• Architected core request routing for a multi-tenant publisher, maintaining zero dropped requests.
• Designed a schema-migration UI that prevents users from breaking their own databases.
• Implemented strict RBAC architecture and defensive data validation.

Training neural networks to track orbital garbage.

• Built a procedural 3D space environment in Blender to generate cheap, synthetic training data.
• Applied Spiking Neural Networks (SNNs) to track high-velocity simulated debris.
• Bridged the sim-to-real gap to stabilize models for physical hardware deployment.

Archive