Computer Engineering Co-op @ University of Guelph (GPA 3.94/4.0) · Cloud Engineer @ Co-Operators · AI Security Researcher
4 merged PRs across pydantic-ai and mcp-atlassian.
| Project | What | Status |
|---|---|---|
| pydantic-ai ⭐ 10k | Fixed Cohere tool_choice silently dropped in v2 chat API. Maintainer extended with VCR tests and named-subset handling |
✅ Merged #5833 |
| mcp-atlassian | Added Confluence permission checking tools | ✅ Merged #1320 |
| mcp-atlassian | Fixed Jira non-image attachments silently dropped by MCP clients (returned as rejected image blob) | ✅ Merged #1469 (co-author, Co-authored-by credit) |
| mcp-atlassian | Wired up cross-request credential validation cache, closing a WAF rate-limit trap | ✅ Merged #1481 (co-author, consolidated with an independently opened fix) |
| pydantic-ai ⭐ 10k | Cohere: implemented request_stream and CohereStreamedResponse, full streaming support where none existed |
🔄 PR #6100, approved, CI green |
| mcp-atlassian | Fixed jira_to_markdown stripping brackets from all bracketed text, not just bare links, silently rewriting [~user] mentions on every read |
🔄 PR #1570 |
| smolagents ⭐ 17k | Fixed KeyError on tool_calls in AmazonBedrockModel with cross-region inference profiles |
🔄 PR #2363 |
| smolagents ⭐ 17k | Fixed broken Space link in tools tutorial (Python 3.13 incompatibility) | 🔄 PR #2364 |
| crewAI ⭐ 31k | Show updated agent answer before feedback prompt when verbose=False |
🔄 PR #6386 |
| Venue | Year | Role | Paper |
|---|---|---|---|
| SIGIR-AP | 2025 | First Author, Best Paper Nominee | FalseCoTQA: Adversarial Multi-Hop QA via Knowledge-Grounded False Chains of Thought |
| CIKM | 2025 | Co-author | Uncovering the Persuasive Fingerprint of LLMs in Jailbreaking Attacks |
| Canadian AI | 2026 | Co-author | Cause-Conditioned Multi-Task Learning for Answerable Question Suggestion in MRC |
| Canadian AI | 2026 | Co-author | From Hints to Answers: Uncertainty-Aware LLM-Guided Retrieval for Multi-Hop QA |
| SIGIR-AP | 2024 | Co-author | UnAnswGen: Generating Unanswerable Questions in MRC |
| Project | Stack | What it is |
|---|---|---|
| Argus | Claude Code, SQLite, FastAPI, React/Vite | Claude Code observability platform. Pre/PostToolUse hooks trace 12+ event types across 29 sessions and 1.4M+ tokens, 198 flagged events tracked, composite trust score |
| AgenticRobot | Claude Code, MCP, FastMCP, Raspberry Pi Pico W | Natural-language robot control via 15 FastMCP tools over TCP WiFi. Mecanum-wheel 4WD, obstacle avoidance, servo radar |
| FalseCoTQA | Python, NLP | SIGIR-AP 2025 Best Paper Nominee. Adversarial multi-hop QA benchmark, knowledge-grounded false chains of thought |
| ResumeWorthy | Next.js, TypeScript, Supabase, Vercel | AI job-application assistant. Resume tailoring, job matching, recruiter finder |
| AKO-Agentic-Kernel-Optimization | Python, Agentic AI, AMD/HPC | Autonomous system that generates, profiles, and refines GPU kernels for high-performance AI inference on AMD hardware |
See the portfolio for more, including the interactive project explorer.
- Co-Operators: Azure and Splunk cloud engineering, IGA/PAM compliance, FastAPI automation across Jira/Confluence
- Cyber Science Lab @ U of Guelph: LLM jailbreak pipelines (97% ASR), RL-based real-time defenses, VLM adversarial vaccination
- OSS: contributing to AI/ML libraries (pydantic-ai, smolagents, crewAI, mcp-atlassian)
Anthropic (2026): Building with Claude API · Claude Code 101 · Claude Code in Action · Intro to Subagents · Intro to Agent Skills · Introduction to MCP · MCP Advanced Topics · Claude 101 · AI Fluency Framework · Introduction to Claude Cowork
Oracle: Agentic AI Certified Foundations Associate
Google Cloud: Vertex AI ML Solutions · ML APIs · BigQuery ML · Security Command Center
Other: Splunk Enterprise Security · MLOps Essentials (NASBA) · Databricks GenAI