Hallucination Detection in LLMs
Compared 8 methods across 200 cases. A grounded approach reached 90.5% accuracy.
- LLM
- Evaluation
- RAG
- Accuracy
- 90.5%
- Methods
- 8
- Cases
- 200
AI EngineerLLM SystemsTrustworthy AI
MATTHEWROCKY
I engineer reliable LLM and agentic systems — from research to production.
- LLM Evaluation
- RAG
- AI Agents
- Production Systems
- Ottawa, Canada
- Open to Work
- M.Eng. (In Progress)
- Research01
- Build02
- Evaluate03
- Deploy04
Scroll
Built with
- Python
- PyTorch
- LangChain
- FastAPI
- React
- AWS
- Docker
01 / Featured Projects
View all projects
Autonomous Taxi AI Ethics
Flags risky drop-off zones using incident reports, severity and recency, with map-based risk overlays.
- AI Ethics
- Risk Maps
- Uncertainty
- Map overlays
- Risk
- Incident signals
- Context
- Risk flags
- Safety
LoopLens
Detects circular public-funding patterns through network graphs, chatbot analysis, and automatic memos.
- NLP
- Graph Analytics
- Agents
- Funding patterns
- Graph
- Data analysis
- Chat
- Auto generation
- Memos
TasteShift Time-Aware Recommendations
Exploring movie recommendations that adapt to changing interests, with explanations and where-to-watch information.
- RECOMMENDER SYSTEMS
- TEMPORAL ML
- PERSONALIZATION
- USER PREFERENCES
- Taste
- CHANGING INTERESTS
- Time
- MOVIE RECOMMENDATIONS
- Discover
IN PROGRESS
Quality-Aware AI Model Router
Exploring intelligent model selection to balance response quality, inference cost, and latency.
- LLM ROUTING
- EVALUATION
- OPTIMIZATION
- RESPONSE EVALUATION
- Quality
- INFERENCE BUDGET
- Cost
- RESPONSE LATENCY
- Speed
IN PROGRESS
02 / About
Engineering judgment,from research to production.
I'm an M.Eng. student in Systems Science & Engineering (Interdisciplinary AI) at the University of Ottawa, with 5+ years in software engineering. My AI/ML work focuses on production LLM systems, retrieval-augmented generation, and trustworthy AI.
More about me
- Experience
- 5+ yearsSoftware engineering
- Location
- Ottawa, Canada
- Education
- M.Eng. (In Progress)University of Ottawa
03 / TOOLKIT
Tools for research.
Built for production.
Languages, frameworks, data systems, and infrastructure used across the portfolio's research and production work.
- RESEARCH
- EXPERIMENT
- EVALUATE
- DEPLOY
AI SYSTEMS
RESEARCH → PRODUCTIONAI / ML
- OpenAILLM APIs
- ClaudeLLM APIs
- GeminiLLM APIs
- LangChainOrchestration
- RAGRetrieval
- AI AgentsAgentic Systems
- Hallucination DetectionTrustworthy AI
- PyTorchDeep Learning
DATA
- PandasData Analysis
- PolarsData Processing
- DuckDBAnalytics DB
- SQLQuerying
- NLPText Processing
BACKEND
- PythonCore Language
- FastAPIAPI Framework
- FlaskWeb Framework
- REST APIsAPI Design
FRONTEND
- ReactUI Library
- Next.jsFull-stack Framework
- TypeScriptTyped JavaScript
- Tailwind CSSStyling
INFRA
- AWSCloud Platform
- DockerContainers
- CI/CDDelivery Pipelines
04 / CONTACT
LET'S BUILDSOMETHINGWORTH SHIPPING.
STATUS / OPEN TO WORK / OTTAWA / REMOTE
Open to remote & hybrid opportunities, collaborations, and interesting problems.