hello...
Nice to meet you! I'm Joseph, an aspiring Machine Learning Engineer with more than 6 years of programming experience.
I am currently on a Co-op term at Darkvision Technologies., an industry leader in NDT pipeline inspection based in North Vancouver.
School-wise, I am pursuing a Bachelor's degree in Computer Science and Physics at the University Of British Columbia, where I am in my third year. When I'm not coding, you may find me making music, bouldering, running, skiing, or immersing myself in the beautiful nature surrounding Vancouver!
my experiences...
Machine Learning Scientist Co-op @ Darkvision Technologies
Sep 2026 - Dec 2026
- • Engineered a computer vision pipeline utilizing RF-DETR, histogram matching, and a novel beamformed view, consistently achieving a 98.5% F1 score across 300km of ultrasound-scanned pipeline data
- • Improved the accuracy of the 6dB defect sizing threshold by 600% through training an XGBoost model to dynamically optimize the threshold based on over 20,000 handpicked image features
- • Spearheaded the integration of several models into production code running on EC2 machines using Kubernetes
- • Built end-to-end pipelines for 3 distinct ML projects, translating stakeholder requirements into deployable models
- • Engineered an ultrasound dataset using hard negative mining and complex augmentations to combat data drift
- • Wrote comprehensive unit tests for production code, and used Teamcity CI/CD pipelines to ensure code quality
Machine Learning Engineer Co-op @ Metrized Inc.
Jan 2025 - Sep 2025
- • Adapted a YOLO detection head to work with a customized DINOv2 backbone to improve the model’s performance on industrial defect detection tasks by over 20%
- • Designed a novel classification algorithm that reduces error rates for text and symbol classification by over 150%using a custom trained mobilenet embedding model with an enhanced KNN algorithm
- • Engineered a cutting edge Image to 3D AI to generate high fidelity models of furniture, helping clients save up to $5000 per unit of shipping and scanning costs.
- • Boosted YOLO object detection speed by 200% by conducting thorough benchmarks and implementing TensorRT.
- • Spearheaded the transformation of a full-stack web app into a distributable desktop app with Electron, rewrote and packaged python backend to use dynamic ports, and bootstrapped large modules to optimize app size
Website Team Lead @ UBC BIOMOD
Oct 2024 - Present
- • Led a team of four to ship a responsive, competition-ready website in less than 3 weeks
- • Finalized the production site by reviewing several pull requests , resolving code conflicts, and fixing critical bugs
- • Architected a well-structured code base and delegated tasks accordingly based on members strengths and interests
- • Constructed an eye-catching home page featuring a controllable 3D model of the team’s microbiology project using Three.js, Tailwind, and React
my projects...
Dance CV - NWHacks 2026 Winner
Mediapipe, Typescript, React, Gemeni API, Vite
Hackathon project that uses computer vision, AI, and voice dictation to help you improve your dance skills
Urbanize AI
AWS Lambda, Bedrock, Expo, React, Python
Hackathon project that promotes sustainable urban development by allowing users to easily visualize improvements with AI
AI Recipe Generator
Next.js, Node.js, React.js, Three.js, OpenAI API
Website that utilizes the OpenAI API to generate recipes based on ingredients found in the fridge entered by the user.
Musical Ear Trainer
Java, Swing, JUnit, SQLite
Helps users practice their interval identification skills by quizing them on randomly generated MIDI notes.
lets talk!
Whether you have an inquiry, want to collaborate on something, or just want to say hi, I'm always willing to chat!
Email me at josephdeng317@gmail.com
hello...
Nice to meet you! I'm Joseph, an aspiring Machine Learning Engineer with more than 6 years of programming experience.
I am currently on a Co-op term at Darkvision Technologies., an industry leader in NDT pipeline inspection based in North Vancouver.
School-wise, I am pursuing a Bachelor's degree in Computer Science and Physics at the University Of British Columbia, where I am in my third year. When I'm not coding, you may find me making music, bouldering, running, skiing, or immersing myself in the beautiful nature surrounding Vancouver!
my experiences...
Machine Learning Scientist Co-op @ Darkvision Technologies
Sep 2026 - Dec 2026
- • Engineered a computer vision pipeline utilizing RF-DETR, histogram matching, and a novel beamformed view, consistently achieving a 98.5% F1 score across 300km of ultrasound-scanned pipeline data
- • Improved the accuracy of the 6dB defect sizing threshold by 600% through training an XGBoost model to dynamically optimize the threshold based on over 20,000 handpicked image features
- • Spearheaded the integration of several models into production code running on EC2 machines using Kubernetes
- • Built end-to-end pipelines for 3 distinct ML projects, translating stakeholder requirements into deployable models
- • Engineered an ultrasound dataset using hard negative mining and complex augmentations to combat data drift
- • Wrote comprehensive unit tests for production code, and used Teamcity CI/CD pipelines to ensure code quality
Machine Learning Engineer Co-op @ Metrized Inc.
Jan 2025 - Sep 2025
- • Adapted a YOLO detection head to work with a customized DINOv2 backbone to improve the model’s performance on industrial defect detection tasks by over 20%
- • Designed a novel classification algorithm that reduces error rates for text and symbol classification by over 150%using a custom trained mobilenet embedding model with an enhanced KNN algorithm
- • Engineered a cutting edge Image to 3D AI to generate high fidelity models of furniture, helping clients save up to $5000 per unit of shipping and scanning costs.
- • Boosted YOLO object detection speed by 200% by conducting thorough benchmarks and implementing TensorRT.
- • Spearheaded the transformation of a full-stack web app into a distributable desktop app with Electron, rewrote and packaged python backend to use dynamic ports, and bootstrapped large modules to optimize app size
Website Team Lead @ UBC BIOMOD
Oct 2024 - Present
- • Led a team of four to ship a responsive, competition-ready website in less than 3 weeks
- • Finalized the production site by reviewing several pull requests , resolving code conflicts, and fixing critical bugs
- • Architected a well-structured code base and delegated tasks accordingly based on members strengths and interests
- • Constructed an eye-catching home page featuring a controllable 3D model of the team’s microbiology project using Three.js, Tailwind, and React
my projects...
Dance CV - NWHacks 2026 Winner
Mediapipe, Typescript, React, Gemeni API, Vite
Hackathon project that uses computer vision, AI, and voice dictation to help you improve your dance skills
Urbanize AI
AWS Lambda, Bedrock, Expo, React, Python
Hackathon project that promotes sustainable urban development by allowing users to easily visualize improvements with AI
AI Recipe Generator
Next.js, Node.js, React.js, Three.js, OpenAI API
Website that utilizes the OpenAI API to generate recipes based on ingredients found in the fridge entered by the user.
Musical Ear Trainer
Java, Swing, JUnit, SQLite
Helps users practice their interval identification skills by quizing them on randomly generated MIDI notes.
lets talk!
Whether you have an inquiry, want to collaborate on something, or just want to say hi, I'm always willing to chat!
Email me at josephdeng317@gmail.com