I build the infrastructure behind production AI applications and turn complex data into reliable decision tools: provider-agnostic SDKs, durable agent runtimes, analytics, and business intelligence dashboards.
My work focuses on making AI systems portable, observable, and reliable across providers. I currently contribute to the Zhivex ecosystem for TypeScript and Python.
Alongside my open-source engineering work, I have delivered data analysis and Power BI dashboards for teams in the banking sector. This work includes data modeling, SQL, Power Query, DAX, KPI design, data quality, and executive and operational reporting.
The source files, datasets, screenshots, client details, and internal metrics are confidential, so I present the capabilities and methods rather than the artifacts.
| Project | Language | Focus | Package |
|---|---|---|---|
| Zhivex AI SDK | TypeScript | Generation, streaming, tools, multimodal AI, provider routing, and durable agents | npm |
| Zhivex AI SDK for Python | Python | Async multi-provider systems, durable state, approvals, safety, tracing, and workflows | PyPI |
| MCP BCRA | TypeScript | Typed MCP access to official financial and banking APIs from Argentina's Central Bank | Repo |
The two Zhivex SDKs expose a consistent multi-provider model while preserving provider-native capabilities. MCP BCRA applies the same reliability principles to a focused, public-data integration.
- Portable agent runtimes with tools, handoffs, approvals, and durable state
- Multi-provider model contracts and explicit native capabilities
- Streaming, structured output, observability, evaluations, and safety
- Data modeling, KPI design, and decision-ready Power BI dashboards
- MCP integrations that turn external APIs into dependable agent tools
- Release engineering and installed-package validation for SDKs
Based in Buenos Aires, Argentina. I work in English and Spanish.