Portfolio
Systems I've shipped to production with measurable impact. Each project bridges research innovation with deployment constraints—because AI that doesn't scale responsibly doesn't matter.
Core research contributions with publications and significant impact.
Novel memory mechanism for coherent LLM-driven dialogue systems. Master's thesis contribution improving long-form context retention through sliding window summarization.
Open-source digital archive search infrastructure for 50,000+ cultural heritage audio recordings. Deployed across 6+ Canadian universities with 61% processing efficiency gains.
Designed and deployed RAG architecture using LangChain and GPT-4. Achieved 21% improved retrieval accuracy and 30% engagement increase serving 5,000+ monthly users.
Research projects with industry applications and academic evaluations.
Led 4-person team developing multilingual conversational AI (6 languages) for social intervention. Novel prompt-driven persona framework with automated UX evaluation.
Systematic evaluation of 7 SOTA models (ResNet, EfficientNet, DenseNet, DeiT, Swin, ViT) on kidney pathology classification with 212K tissue images.
Edge-optimized computer vision pipeline for classroom engagement detection. Achieved 30 FPS on resource-constrained devices with privacy-preserving approaches.
Production deployments with measurable business outcomes.
Modernized digital humanities infrastructure with Docker containerization and CI/CD pipelines. Reduced deployment time by 80% and improved processing efficiency by 61%.
Real-time image search engine using TensorFlow on AWS with 16% false positive reduction. Built fraud detection pipelines achieving 92% accuracy on 1M+ daily transactions.
Developed recommendation system using scikit-learn for ProfCess, improving content relevance by 40%. Enhanced frontend performance across core user flows.
Experimental systems and academic explorations.
Web application using Gaussian Naive Bayes to predict job descriptions from candidate profiles. Full-stack ML prototype with interactive interface.
Image processing and visualization pipelines analyzing crop yield signals from satellite data using Google Earth Engine.
Technical review and presentation on quantum computing applications in computational immortality and consciousness modeling.
Core technologies and frameworks across my work.
PyTorch, TensorFlow, Hugging Face Transformers, LangChain, LlamaIndex, LangGraph, Vector Databases (Pinecone, Weaviate)
GPT-4, Claude, Gemini, Llama, Mistral with fine-tuning, prompt engineering, and RLHF experience
AWS (EC2, Lambda, S3, SageMaker), Azure AI, Google Cloud, Docker, Kubernetes, GitHub Actions CI/CD
Apache Spark, PySpark, Databricks, Apache Airflow, PostgreSQL, MongoDB, Redis, ETL Pipelines