Wenhe Wang

Hello! 👋

I'm Wenhe Wang, a Data Scientist with a Master's in Visual Computing from SFU. I specialize in computer vision, NLP, and cloud data solutions, and have experience developing models to optimize fraud detection, process image, enhance data pipelines, and automate text analysis.

My work spans sectors from CAD, BIM, Edge Devices, FinTech, Consulting to Short Video industry, including leading projects at Autodesk, Samsung Research Canada, Cavallo Technologies, TikTok, China Guangfa Bank, and KPMG.

I received my M.S. in Financial Engineering from University of Reading in 2019 and my B.Ec. in Financial Engineering from Nankai University in 2017.

Projects

These are some selected personal and course projects that I have worked on, throughout my studies.

Unity Cyberpunk Hologram and Teleport effects

Wenhe Wang and Gouttham Nambirajam
SFU Class Final Project, CMPT 766: Computer Animation

Final Effect, Presentation Slides

This is a from scratch re-implementation with Unity Shader script and Unity Rendering Pipeline to realise two 3D visual effects.

Preference Optimization

Wenhe Wang, Gouttham Nambirajam, and Raj Shreyas Penukonda
SFU Class Project, CMPT 713: Natural Language Processing

Project Page

Implemented preference optimization using the ORPO framework to fine-tune the Qwen 2.5 0.5B-Instruct model for instruction-following tasks.

Prompt Tuning for Text Generation

Wenhe Wang, Gouttham Nambirajam, and Raj Shreyas Penukonda
SFU Class Project, CMPT 713: Natural Language Processing

Project Page

Fine-tuned GPT2 using prompt tuning techniques to tackle table-to-text generation tasks.

BERT Finetuning for Robust Phrasal Chunking

Wenhe Wang, Gouttham Nambirajam, and Raj Shreyas Penukonda
SFU Class Project, CMPT 713: Natural Language Processing

Project Page

Fine-tuned a BERT-based Transformer model to tackle phrasal chunking.

Word Vectors and the Analogy Task

Wenhe Wang, Gouttham Nambirajam, and Raj Shreyas Penukonda
SFU Class Project, CMPT 713: Natural Language Processing

Project Page

Optimized word vectors to solve analogy tasks, explored the use of retrofitting techniques to enhance the semantic relationships captured by pre-trained GloVe embeddings.

Segmentation into Words

Wenhe Wang, Gouttham Nambirajam, and Raj Shreyas Penukonda
SFU Class Project, CMPT 713: Natural Language Processing

Project Page

Automatically segment English input that does not have word boundaries into the most plausible sequence of words with N-gram.

3D Reconstruction using Neural Fields (NeRF)

Wenhe Wang
SFU Class Project, CMPT 743: Practices in Visual Computing 2

Project Page

Implemented a neural network for 3D object reconstruction from occupancy field samples, leveraging a 3D feature grid and a Multilayer Perceptron (MLP).

Latent Manipulation with StyleGAN2

Wenhe Wang
SFU Class Project, CMPT 743: Practices in Visual Computing 2

Project Page

Delved into various aspects of StyleGAN2, including basic image generation, latent space interpolation, and advanced manipulation techniques like latent mixing, showcasing the model's capability to blend and alter image features seamlessly.

Image to image translation with CycleGAN

Wenhe Wang
SFU Class Project, CMPT 743: Practices in Visual Computing 2

Project Page

Re-implemented CycleGAN from scratch to convert images from one domain (e.g., Apple emojis) to another (e.g., Windows emojis) without direct pairings, maintaining the intrinsic structure of the original images.

Pose Regression with PoseNet

Wenhe Wang
SFU Class Project, CMPT 743: Practices in Visual Computing 2

Project Page

Re-implemented PoseNet with modified InceptionV1 architecture pre-trained on the Places dataset for estimating the camera's position and orientation and pose estimation from single images.

Object Detection with SSD

Wenhe Wang
SFU Class Project, CMPT 743: Practices in Visual Computing 2

Project Page

Re-implemented a simplified version of the Single Shot MultiBox Detector (SSD) for object detection, diverging from the conventional YOLO approach by incorporating multiple scales and default bounding boxes as anchors.