Harness Engineering
View allMy npm skill pack — install once, then agents follow scope, verify, and resume workflows.
ScaffoldAuditStateLifecycleScopeVerification
Solano de Zuasnabar
AI Engineer
Tucumán — Argentina
AI Engineer with experience building and deploying production-grade AI systems — from LLM-powered backends and RAG pipelines to computer vision and automation workflows. I work autonomously in fast-moving environments, taking ownership from prototype to production and iterating based on real usage and stakeholder feedback.
Experience
View all
Tecnología BI· AI EngineerNow
Design and deployment of production AI systems for multiple clients — direct stakeholder contact and ownership from prototype through iteration in production.
Blue Star Group · Isadora & Todomoda
Built and maintained AI tools for a large fashion retail group — semantic product search with store maps, product description workflows, and a multimodal design assistant for internal teams.
Integra · Agricultural SaaS
Owned two FastAPI backends for an agricultural SaaS product — a natural-language analytics chatbot over warehouse views, and a read-only alerts platform with cached snapshots and optional AI-assisted discovery.
Develative· AI Engineer
Vento Ventures · Venture capital firm
Venture capital startup. Built AI-driven automation workflows to support an internal investment management platform.
How I work with AI
Cursor agent skills I reach for daily — planning, debugging, and shipping with structure.
I built Harness Engineering — open-source agent skills for Cursor, Claude Code, and Codex, based on Learn Harness Engineering.
npx harness-skills installView repo
Latest writing
View all- Observability for AI Systems: Tracing, Evals, and Feedback Loops
AI-powered systems need a different kind of observability than traditional software. This guide covers tracing LLM calls, tracking evaluation scores over time, and building feedback loops that let you improve continuously.
AI engineeringObservabilityLLMTracingProduction - How AI Tools Actually Work
Models do not read files or run commands on their own. They ask for tools, and a harness runs them.
AIToolsAgentsTooling - LLM Evaluation in Practice: Beyond MMLU
MMLU tells you nothing about whether your LLM-powered application works. This is a practical guide to evaluating LLM pipelines for production — harness-based benchmarks, LLM-as-judge, and the signals that actually matter.
LLMEvaluationBenchmarksAI engineeringQuality
Stack & Tools
Languages
PythonTypeScriptJavaScript
AI & LLMs
OpenAI APIGoogle AIRAG pipelinesEmbeddingsSemantic searchMultimodal searchTool calling & agents
Backend
FastAPICeleryREST & webhooksSSE streamingPydanticPolarsWeasyPrint
Data & storage
SQL ServerPostgreSQL / SupabaseRedisQdrantPinecone
Frontend
Next.jsStreamlitTailwind CSSshadcn/ui
Automation & infra
Dockern8nZapierGit / GitHubVercelMCP servers
Also worked on
Harness Engineering skills. npm package (harness-skills) with agent skills for Cursor, Claude Code, and Codex — npx harness-skills install.
RAG chatbots end-to-end. Document ingestion, chunking, embedding management, vector store integration, and LLM response generation.
Web scraping pipelines. Data collection and enrichment with BeautifulSoup and requests.
MCP server integrations. Connecting AI models with external tools and data sources.
Supabase-backed projects. Personal projects using PostgreSQL for backend data management.
Languages
Spanish — Native
English — Advanced — C1