AI Agents & LLMs
Conversational AI agents, LLM integrations, prompt engineering, semantic search, and RAG — built with production-grade guardrails, not demos.
Welcome
Gastón Isnardi
Backend · Cloud · AI Engineer
I build production AI systems and scalable serverless backends — conversational AI agents, high-volume financial platforms, and cloud infrastructure. I work with companies of every size, from startups to enterprises, across Latin America and beyond. Things that work at scale, not just in demos.
- Node.js
- TypeScript
- AWS
- Serverless
- LLMs
- Next.js
What I do
Services
I help companies of every size — from early-stage startups to large enterprises — ship production-grade AI, backend, and cloud systems. Need something built? Let me know.
Backend & Serverless APIs
Scalable backends and APIs on AWS — Lambda, event-driven architectures, and serverless platforms that handle real load.
Cloud Architecture & AWS
Infrastructure design, IaC with AWS CDK, cost optimization, authentication, and contact-center integrations end to end.
Full-Stack Web Development
Modern web apps with Next.js and React, from idea to production deployment — fast, accessible, and SEO-ready.
Featured Work
Projects & Case Studies
Production AI systems and serverless infrastructure built for enterprise clients, plus side projects shipped to production.
Enterprise work
Retail AI Shopping Advisor
2025Zero-hallucination production prompt
Conversational AI agent for WhatsApp that guides users through purchasing computers and tablets. It identifies needs, runs semantic search over a product catalog, validates data, and presents recommendations via interactive cards — built with strict honesty rules and zero-hallucination guardrails in a production prompt.
LLMsPrompt EngineeringSemantic SearchVector DatabasesNode.jsTypeScriptWhatsApp Business API
Yalo
Campanify — Multichannel Campaigns Platform
2024Hybrid-RDS cost optimization
Serverless platform (ConfA) to manage and run multichannel marketing campaigns — email, SMS, push — with audience segmentation from spreadsheets. A fleet of Lambda functions with an SQS → SES pipeline, plus a hybrid RDS Aurora Serverless architecture (shared dev/qa, dedicated prod) that cut infrastructure cost significantly.
AWS LambdaAWS CDKTypeScriptPrismaRDS Aurora ServerlessSQSSESNext.jsReact
Cloudhesive
Citikon — Digital Civil Registry
2024Granular authz with Cedar policies
Full-stack serverless platform digitizing civil-registry and corporate-justice procedures for Argentine government bodies. Domain-organized Lambda functions and granular per-endpoint authorization via AWS Verified Permissions with Cedar policies, plus admin and citizen Next.js portals authenticated through Cognito.
AWS LambdaServerless FrameworkTypeScriptPrismaAWS Verified PermissionsCedarAmazon CognitoNext.jsMUI
Cloudhesive
Own projects
Prode Mundial 2026
2026Zero-cost production stack
Web app for World Cup 2026 match predictions, running at zero monthly cost. Built with Next.js + Supabase (Postgres) + Prisma on Vercel, with Postgres RLS policies, scoring logic, and a Docker-based local dev setup.
Next.jsSupabasePostgreSQLPrismaVercelTypeScript
Personal projectLive ↗
Mes del Trabajador — Working&Co
2026Resilient production landing
Landing page with an interactive events calendar, program pillars, and a raffle for a large internal audience. Built with Next.js + TailwindCSS on Vercel, with a resilient data layer and graceful fallback to local data.
Next.jsReactTailwindCSSVercelTypeScript
Personal projectLive ↗
Experience
Work History
Software Engineer
Yalo
Nov 2025 — Present
Remote
Design, build, and optimize production conversational AI agents for WhatsApp, serving enterprise clients in retail, foodservice, and airlines across Latin America. Work spans a retail shopping advisor with semantic product search and a zero-hallucination production prompt; a multi-agent foodservice system (Coordinator + specialized Worker) with promotion search and systematic failure analysis; and a dual-role airline agent (upsell + travel-policy support) with purchase-intent detection and handoff to human agents. Responsibilities include advanced LLM prompt engineering (guardrails, frustration detection, ambiguity handling, intent routing), semantic search integration with vector stores over GCP Cloud Functions and the WhatsApp Business API, custom Lua tools for analytics tracking and interactive product cards, systematic agent QA, and technical documentation.
Node.jsTypeScriptLLMsPrompt EngineeringLuaVector StoresWhatsApp Business APIGCP Cloud Functions
Full Stack Developer
Cloudhesive
Jun 2022 — Nov 2025
Buenos Aires · Remote
Worked closely with development teams to design, build, test, and ship solutions with React, Angular, Node.js, and AWS, leading projects for clients across education, telecom, government, finance, and media. Highlights: education platforms for Universidad de Los Lagos, Agencia de Calidad de la Educación, and Universidad Andrés Bello (student datasets, enrollment, and registration apps with React + AWS); a full dashboard for Banco Estado (Angular + Node.js + AWS CDK); an AWS Connect contact center for Comfenalco Quindío (CDK, DynamoDB OTP, call-triggered Lambdas, WhatsApp integration, GitHub Actions deploys); a digital marketing email service for Enersa (CDK, Node.js, SNS, SES, DynamoDB); and a serverless Slack bot for Infobae using Lambda, DynamoDB, and Amazon Bedrock. Also designed hybrid-RDS cost optimizations and granular authorization with AWS Verified Permissions.
Node.jsTypeScriptAWS LambdaAWS CDKDynamoDBRDS Aurora ServerlessPrismaSQSSESCognitoNext.jsGitHub Actions
Full Stack Developer
BesySoft S.A.
Dec 2020 — Jun 2022
Buenos Aires
Developed enterprise banking systems for Banco Hipotecario (ProCreAr) and Banco San Juan using Java and Angular. Contributed to high-security financial applications in agile teams.
JavaAngularMySQL
Let's build something
Have a project or an idea to build? I work with companies of every size on backend, cloud, and AI. Let's talk.