Argentina
I’m a software engineer based in Buenos Aires, Argentina and I love helping craft new products. I have over 20 years of experience as a Software Engineer developing applications for big corporations and startups, wearing several hats, always with a positive mindset and the passion to solve problems and help others as well. What else? I am passionate about roles that bring a lot of very interesting design challenges and enjoy working with a team of talented engineers to create innovative and user-friendly applications.
AI Engineering — [Dental Health AI Company] — Built a production-grade knowledge assistant indexing 100K+ Slack messages and 500+ Confluence pages, enabling semantic search across organizational knowledge. — Designed a hybrid retrieval architecture combining BM25 sparse vectors with dense semantic embeddings, using Reciprocal Rank Fusion to merge result sets and cross-encoder reranking to surface the most relevant documents — significantly outperforming pure semantic search on domain-specific queries. — Built an evaluation harness using the LLM-as-judge pattern — Claude scoring pipeline responses against a curated ground truth dataset — achieving 85% accuracy. Used Arize Phoenix for end-to-end observability and tracing, keeping precision and recall visible across live queries to prevent prompt overfitting. AI Engineering — [Consumer goods company] — Built a Post-meeting Intelligence pipeline that transformed raw, unattributed transcripts into structured intelligence: action items, decisions, open questions, and speaker ownership — with no prior speaker metadata available. — Solved the attribution problem with a dedicated speaker inference prompt, effectively creating a two-pass architecture where identity resolution preceded content extraction. — Enforced extraction boundaries using Pydantic-based structured output schemas and explicit negative constraint instructions, ensuring the model only surfaced what was explicitly stated rather than inferred or hallucinated. — Validated output quality using an LLM-as-judge evaluation harness, achieving a sub-5% error rate across all extraction tasks against a curated ground truth set.
Working as a founding engineer for an startup company that's building a Fantasy Football app. I've designed the application architecture, and also implemented features I've used Next.js because it was the perfect framework for this type of product and used a document-based DB also since it's very flexible. The product gives the user to ability to use PayPal, Venmo and Credit Cards to fund their Wallet.
As a Full Stack Engineer at BetterEngineer, I played a key role in enhancing a platform that enables users to earn automatic cashback on purchases made with participating merchants by linking their credit or debit cards to the app. The challenge was to augment the platform with new features, adding reports like Year-Over-Year sales reports and customer segmentation tools to provide valuable insights for our merchants. I was an active maintainer of 4 projects, and not only added the desired features, but also gained considerable experience using Terraform as our infrastructure-as-code solution, and CircleCI as our CI/CD solution. The introduction of these new features has significantly improved the user experience, providing merchants with crucial data insights, and ultimately driving increased usage of the platform.
This project entailed designing a state-of-the-art social network for live video streaming at Genium. The platform was built using Next.js, GraphQL, and WebSockets, and was deployed on AWS EC2, enabling over 10,000 users to share their daily activities and foster stronger connections within their communities. The challenge was to create a platform that supports seamless real-time interactions and enhances user engagement in a social setting with live video and real-time chat. As the Software Architect, I leveraged my expertise in React, PostgreSQL, JavaScript, TypeScript, and Node.js to develop the platform. I also utilized Nest.js framework that greatly helped to get the backend running in just a few weeks