Prysai Lab · Repositories · Research trail
I explore how AI agents move from impressive demos to useful, inspectable systems in real developer workflows.
My current focus:
- agent tooling and local developer workflows
- RAG evaluation, tracing, and regression testing
- MCP, browser automation, and tool-using agents
- Codex skills, task design, and evidence-driven delivery
I am studying the full agent engineering loop: choose the right model, provide the right context, connect tools safely, evaluate behavior, and turn successful experiments into repeatable workflows.
The goal is simple: fewer impressive demos, more systems that can be tested, understood, and improved.
Prysai is the home for my ongoing product experiments, research notes, and practical work around AI and developer tools.
The repositories below are working copies and study paths for upstream projects I am exploring or contributing to:
| Area | Reading list |
|---|---|
| Agent frameworks | Google ADK · PydanticAI · FastMCP |
| RAG and evaluation | RAGOps · Opik · Promptfoo |
| Developer workflows | Jupyter AI · Browser Use · Spec Kitty |
These are intentionally presented as research and contribution paths, not as original projects.
What I am exploring
| Agent systems How agents use context, tools, memory, and feedback. |
RAG and evaluation How to test retrieval, quality, regressions, and failure modes. |
| Developer experience How local workflows make agent work easier to inspect and repeat. |
Codex workflows How tasks become bounded, verifiable, and useful in practice. |
This profile includes a mix of experiments, learning repositories, and forks used for study and upstream contribution. The repository list is part of an active research process—not a claim that every repository is an original project.
AI Agents · RAG · LLM Evaluation · MCP · Developer Tools · Codex Workflows
Building in public, one reliable workflow at a time.