Research Projects Skills Contact Get in touch
Available for research collaborations · July 2026

AI / Machine Learning Engineer & Computer Vision Specialist

Specializing in efficient deep learning, computer vision pipelines, generative architectures, and edge optimization — turning research-grade models into production systems that run fast, scale cleanly, and ship reliably.

View Projects Read Research Contact Me
GitHub trancongman276@gmail.com
Featured Research
01 / RESEARCH
Paper in Progress Lead Researcher & Architect

Generative AI for 3D Sequence Recognition

A pioneer encoder–decoder framework that fuses Temporal Convolutional Networks with Mamba-v2 State Space Models for continuous sequence synthesis and spatial–temporal pattern recognition — outperforming RNN/Transformer baselines on long-sequence modeling while reducing inference complexity.

20,000+
trajectory samples
~2,000
fine-grained labels
10+
contributors
Key Contributions
  • Novel Generative Architecture
    Encoder–Decoder integrating TCN + Mamba-v2 for continuous synthesis.
  • Large-Scale Benchmark Dataset
    Curated continuous trajectory dataset with 20k+ samples & 2k labels.
  • Performance Gains
    Outperformed RNNs/Transformers on long-sequence spatial modeling.
Mamba-v2 TCN State Space Models PyTorch Sequence Modeling
Engineering Projects
02 / WORK

Production systems across edge vision, analytics, and multimodal AI.

A selection of anonymized engagements — architecture, optimization, and deployment across constrained and distributed environments.

PROJECT

Real-Time Edge Vision & Traffic Analytics

Edge
Computer Vision & Inference Optimization Engineer
  • ▹ Optimized DL inference with YOLO, ByteTrack, TensorRT/ONNX — ~50% FPS uplift on constrained hardware.
  • ▹ Built multi-camera pipelines for tracking, entry/exit, speed estimation & plate recognition via OpenCV.
  • ▹ Automated QA & edge-case suites for tracking stability under occlusion.
PyTorch OpenCV TensorRT ONNX ByteTrack YOLO
PROJECT

Modular Smart Analytics Engine

Platform
Lead AI Systems Architect
  • ▹ Architected an end-to-end modular platform with plug-and-play inference modules, flexible DB bindings, and decoupled async queues.
  • ▹ Evaluated SOTA vision models to optimize precision ↔ hardware overhead trade-offs.
  • ▹ Engineered internal CLI tools & automated debugging suites for client deployment.
System Architecture Python REST APIs Microservices Docker
PROJECT

Vision-LLM Automated Spatial Quantification

Multimodal
AI Solutions Researcher
  • ▹ Generative vision pre-processing via multimodal APIs (Gemini AI) + advanced prompt engineering to clean blueprint noise while preserving structural lines.
  • ▹ Combined server-side OCR with classical OpenCV geometric algorithms to automate floor & spatial area calculation.
Multimodal LLMs Prompt Engineering OpenCV OCR Image Processing
PROJECT

Cross-Platform Inference & Dataset Curation

Cross-platform
ML / Optimization Engineer
  • ▹ System audits & ONNX Runtime performance profiling across desktop environments.
  • ▹ Feasibility studies for mobile hardware compilation (iOS / CoreML).
  • ▹ Curated high-precision annotations for modern detection architectures (YOLOv9-DEYO).
ONNX Runtime CoreML Feasibility Data Engineering YOLOv9
Technical Skills
03 / STACK

A focused stack for research-to-production AI.

Core AI & DL

PyTorch Mamba-v2 TCN Vision LLMs Prompt Engineering YOLO

Computer Vision

OpenCV ByteTrack Object Tracking Image Segmentation Spatial Analysis

Inference & Optimization

TensorRT ONNX Runtime Quantization Edge Optimization

Systems & Tools

Python C++ Docker Git CI/CD System Architecture