I build AI agents and the cloud tooling around them at Hewlett Packard Enterprise (Hybrid Cloud & Office of the CTO, Galway), and I'm in my final year of BSc (Hons) Software Design with AI for Cloud Computing at the Technological University of the Shannon (TUS).
The part I enjoy most is making AI tools that people can actually rely on. My final-year project looks at the same problem from the safety side.
- Work: Innovation Engineer (part-time) at HPE (Sep 2026 - Present), joining full-time in June 2027 after completing my internship (Jan - Aug 2026).
- Final-Year Project: Building an open-source evaluation tool on Inspect AI and AgentHarm that separates safety refusals from capability failures in small tool-calling agents.
- Leadership: Chair of the RUN-EU Student Council, representing over 15,000 students across nine European universities.
- Certification: AWS Certified AI Practitioner (Jul 2026)
Most of my recent work is internal at HPE (written up in detail on cgallagher.dev/projects):
- Knowledge Agent for a Multi-Agent Platform (HPE): An agent other platform agents query for product docs and asset data via A2A, MCP, and pgvector. Answers are assembled in code rather than free-written by the model. Fastest PoC on the team to reach the product roadmap; led to a patent mining session and two co-authored papers (HPE Innovation Days & TechCon).
- AI Report Generation Tool (HPE): Turns customer hardware spreadsheets into validated Word, PowerPoint, PDF, and Excel reports using Pandas, Pydantic, and an LLM narrative layer. Led the architecture and backend from empty repo to production approval on Kubernetes in ten weeks (HPE Bronze Star Award).
- Agentic Cloud Migration Platform (HPE): Built the storage and pre-flight check action agents, and helped build the conference demo UI with teams in Ireland, India, and the US.
| Project | Description | Tech Stack |
|---|---|---|
| Small Agent Safety Evaluation (in progress) | Open-source evaluation tool that pairs harmful and benign agent tasks using the same tools to separate safety refusals from capability failures across small local models. | Python, Inspect AI, AgentHarm, Ollama, pytest |
| bee-safe-cv | On-device mobile vision app built during a RUN-EU programme in Belgium to detect Varroa mites on honeybees offline. | Python, YOLOv8, TFLite, Flutter |
| Distributed Inventory Manager | Role-based CLI inventory and sales system built across team sprints with a full CI/CD and static analysis pipeline. | Python, SQLite, pytest, Jenkins, SonarQube |
| Mushroom Classifier | Explainable decision-tree classifier for predicting mushroom toxicity. | Python, scikit-learn, Pandas, PyQt5 |
- Portfolio: cgallagher.dev
- LinkedIn: linkedin.com/in/charliegallagher2004
- Email: cgallagher.dev@gmail.com