Applied AI engineer, OSS maintainer, founder of Kumanday Intelligence, and VP of Trilogy AI Center of Excellence.
I build applied AI systems at the intersection of agentic engineering, work orchestration, evaluation, multimodal generation, and adaptive education. My current work focuses on turning frontier AI capabilities into practical systems that help engineers, researchers, creators, and learners do better work.
- Work orchestration for agentic engineering
- AI-assisted software development workflows
- Context engineering, memory, and repository intelligence
- Multi-agent systems for research, coding, and operations
- LLM evaluation, LLM-as-judge methods, and evaluation-driven development
- Retrieval-augmented generation, GraphRAG, and grounded knowledge systems
- Multimodal media generation pipelines
- AI-powered education and technical training for Latin America
| Project | Focus | Links |
|---|---|---|
| OpenSymphony | Work orchestration for agentic engineering: work units, memory capsules, repo context, human review loops, and bidirectional integration with agent harnesses. | https://opensymphony.dev • https://github.com/kumanday/OpenSymphony |
| cli_engineer | Rust-based experimental agent harness for autonomous coding, code review, repository analysis, and multi-provider LLM workflows. | https://cli.engineer • https://github.com/trilogy-group/cli_engineer |
| Nexus Agents | Multi-agent deep research system with role-based coordination, tool use, human-in-the-loop review, and persistent knowledge base integration. | https://github.com/trilogy-group/nexus-agents |
| G-Eval / Opik Suite | Reference-free LLM-as-judge evaluation framework with structured metrics and repeatable experiment workflows. | https://github.com/trilogy-group/ai-coe-opik |
| GraphRAG POC | Graph-based retrieval-augmented generation prototype using dynamic knowledge graphs and hybrid retrieval. | https://github.com/trilogy-group/ai-coe-graphrag-poc |
| TTV Pipeline | Text-to-video generation pipeline for segment planning, keyframes, chaining modes, and cloud/local generation backends. | https://github.com/trilogy-group/ttv-pipeline |
| media-tooling | Utilities for automated image, video, and audio generation workflows, including provider adapters, batch generation, prompt templates, and asset pipelines. | https://github.com/kumanday/media-tooling |
- Agent runtimes and orchestration protocols
- Model Context Protocol integrations
- AI coding agents and software delivery automation
- RAG systems that preserve provenance and grounding
- Evaluation harnesses for LLM applications and agents
- Developer tools that make AI workflows auditable and repeatable
- Multimodal pipelines for synthetic media creation
- Adaptive learning systems and AI education platforms
I publish, teach, and mentor around applied AI engineering, especially for builders working across Latin America.
Recent themes include:
- Reference-free LLM evaluation with G-Eval
- Designing multi-agent systems for deep research
- Work orchestration for agentic engineering
- RAG, GraphRAG, and context engineering
- AI coding agents and engineering productivity
- Multimodal AI systems
- Paper-to-code AI engineering education
- Website: https://kumandayintelligence.com/
- LinkedIn: https://www.linkedin.com/in/leonardo-gonzalez-technology
- YouTube: https://www.youtube.com/@leonardo-gonzalez-technology
- X: https://x.com/archimagos
- GitHub: https://github.com/kumanday
- Trilogy AI Center of Excellence: https://ai-coe.trilogy.com/
- Technical blog: https://trilogyai.substack.com/