Collin A. P. Joseph, MEng.

• Frontend application development using Angular.
• Upgraded production application from Angular v19 to v22.
• Accelerated development using GitHub Copilot.

• Published and maintained multi-platform (.NET Core & .NET Framework) C# application.
• Built C# REST API client to upload 3D models to cloud rendering platform.
• Deployed C++ tool to generate and manipulate HDR image-based lighting environment maps.
• Led multi-national development team while continuing hands-on development.
• Shipped 12 features across 3 key software products per year using Agile Project Management.

• Delivered new features for Computer Vision inference server using C++.
• Prototyped new CNN models and updated existing ones using Tensorflow.
• Automated data pre-processing pipeline and model benchmarking using Python.
• Deployed watchdog client using Python for server monitoring and control.
• Shipped new features and updates to clients using Docker.

• Developed ensemble classifier for medical microwave radar scans using Python.
• Performed extensive statistical analysis on radar scans using MATLAB and R.
• Optimized ensemble training using advanced adaptive algorithms.
• Improved predictive performance by proposing new time-frequency decomposition features.
• Streamlined data pre-processing by developing tools in Python.

• Prototyped an environment classification algorithm for low-resource digital hearing-aids (MATLAB & Assembly).
• Optimized acoustic noise reduction algorithm in Assembly, decreasing power consumption by 20%.
• Leveraged fixed point arithmetic operations to maximize algorithm efficiency on embedded hardware.
• Executed rigorous regression testing on algorithm to ensure performance.

• Prototyped a digital equalizer firmware module in Assembly Language.
• Simulated and performed experimental analysis on digital equalizer using MATLAB.
• Executed and documented extensive test procedures for acoustic feedback cancellation algorithm.
• Documented and presented analysis and test results to firmware and software development team.

• Prototyped a digital equalizer firmware module in Assembly Language.
• Simulated and performed experimental analysis on digital equalizer using MATLAB.
• Executed and documented extensive test procedures for acoustic feedback cancellation algorithm.
• Documented and presented analysis and test results to firmware and software development team.

• Tutored engineering students in C# programming fundamentals.
• Presented course content to students on an individual basis and in classes of up to 100.
• Consistently completed administrative and pedagogical tasks ahead of deadlines.
• Cooperated effectively with teaching team of 18 members to ensure smooth running of the course.

• Developed and evaluated a ray tracing simulation tool for radio wave propagation using C++ and MATLAB.
• Improved runtime by over 60% using CUDA parallel computing platform on Nvidia GPU hardware.
• Built GUI using MATLAB to visualize simulation results.

• Installed domestic and commercial scale photovoltaic power systems.
• Assisted in project layout calculations and design.
• Streamlined project cost estimation using Microsoft Excel.

• Relevant coursework: Machine Learning, Deep Learning, Computer Vision, Natural Language Processing
• GPA: 4.0/4.0
• Honors: President's List

• Relevant coursework: Machine Learning, Signal Processing, Statistical Analysis
• Thesis: Statistical analysis and machine learning algorithms for RF breast cancer screening

• Co-op program (24 months of industry internships in total).

• CAPE (A-level) Physics, Mathematics & Chemistry

Stack: Python, Langchain, ChromaDB, GeminiAPI
Repo: github/personalized-medication-assistant

Smart assistant that uses GeminiAPI and Retrieval Augmented Generation to provide information about user medication regimens. Everyday users can use this intelligent assistant to quickly obtain relevant information about medications stored in their profile.

Stack: Python, FastAPI, Docker, GCP, PyTorch, Hugging Face
Repo: github/text-from-prescriptions

Containerized REST API that identifies medication brand names from images of handwritten text.
API can be used to build cloud-deployable web applications to efficiently digitize physical medical records.

Stack: Python, PyTorch, Hugging Face
Repo: github/aig230-project

NLP pipeline which classifies review sentiments using BERT base model and custom classification head.
Project demonstrates effective application of pre-trained transformer models to custom machine learning tasks.

Stack: Python, Media Pipe, Scikit-Learn
Repo: github/aig210-project

Computer vision pipeline using Media Pipe and a custom classification head to recognize hand gestures in photos.
Project demonstrates effective use of existing machine learning libraries and with custom model development.

Stack: Python, FastAPI, PyTorch, Docker, GCP
Repo: github/handwritten-digit-api

Containerized REST API that accepts handwritten digits and returns classifications using a CNN model.
Project demonstrates effective use of data science frameworks, containerization and cloud-deployment tools.