Leo Rojas
This site is a small collection of working prototypes I’ve built to explore how modern technologies (including AI) can be applied in practical, responsible ways.
As an engineering and software quality leader, I use projects like these to understand new tools end-to-end before adopting them. Rather than starting with theory or vendor claims, I prefer to validate ideas by building something real, even if it’s intentionally simple.
Each project on this site is designed to:
- Explore a specific technical pattern or capability
- Reduce ambiguity around new technologies
- Serve as a concrete example I can use to guide and inspire teams
- Demonstrate how small, focused proofs of concept can lead to better decisions at scale
The emphasis here is not polish or completeness — it’s clarity, learning, and leadership through execution.
Projects
Single-Agent Testing Loop With Tool Use — This proof of concept focuses on the foundational agentic pattern: an LLM equipped with tools that can explore a live UI, determine next actions, evaluate responses, and adapt in real time instead of executing fixed automation scripts. The current version is intentionally scoped as a small single-agent implementation to validate the approach. A production implementation would likely introduce persistent memory, multi-agent task delegation, and tighter CI/CD integration.
RAG Chatbot — Tool for live testing of non-deterministic chatbot responses and running a small human-reviewed regression suite against known ground-truth answers
Netscape — A lightweight proof of concept exploring how large language models can be integrated into applications via API.
Tip Splitter — A small application demonstrating how clear business rules and AI-assisted development can be used to build practical internal tools.
painting force — A real-world service business website demonstrating platform selection, SEO-first design, and paid acquisition to deliver measurable outcomes without over-engineering.
flowise to automate testcases from requirements An agentic AI built with Flowise that converts plain-English software requirements into structured, automated test cases. This proof-of-concept is a hands-on, empirical demo of agentic AI capabilities—real-world deployment would require far more robust design, validation, and infrastructure.
Each project includes a short case study describing the architecture, intent, and how similar patterns could be applied by an engineering or SQA team in a real organization.