Full Stack Developer

Overview

Since 2022, has worked at Unosquare as a Full Stack Developer across two major engagements, both with substantial LLM integration. Currently on Harvard University's online learning platform: builds and maintains scalable services in a Vue 3, Nuxt 3, and NestJS microservices architecture, and architected a Claude-powered agent that answers internal natural-language questions — "which cohort is learner X in?", "how many learners are in cohort Y?" — by translating prompts into safe database queries via tool use. Earlier, on the Behavox behavioral-analytics platform, designed a Retrieval-Augmented Generation (RAG) chatbot over the company's internal knowledge base, giving employees grounded answers over policies and technical documentation.

Tech deep dive

On the Harvard engagement: Vue 3, Nuxt 3, NestJS, MikroORM, PostgreSQL, Redis, gRPC, and AWS in a microservices architecture, with a strong emphasis on legacy modernization, API performance, and LLM integration. The natural-language query agent uses the Anthropic Claude API with function/tool calling to expose a curated set of read operations, structured outputs for reliable parsing, and prompt engineering with guardrails to keep responses grounded and in-scope; the internal dev workflow consumes existing MCP (Model Context Protocol) servers, and spec-driven development with Claude Code turns written specs into scaffolded views and services. On the Behavox engagement: React, Next.js, and TypeScript on the front end, NestJS microservices integrated with external APIs, AWS API Gateway for security, and CRON-based workflow automation; the RAG chatbot combined vector embeddings, semantic retrieval, prompt templates, and evaluation loops across Anthropic and OpenAI models to keep answers grounded in the internal knowledge base.

Key achievements

• •Architected and implemented a Claude-powered, tool-using chatbot on Harvard's platform that answers internal natural-language questions against the database (learner-to-cohort lookups, cohort membership counts) via function calling and safe query mediation.
• •Introduced spec-driven development with Claude Code and MCP servers on the Harvard engagement, cutting the build time of new views from ~1 week to ~3 days.
• •On the Behavox engagement, designed and shipped a Retrieval-Augmented Generation chatbot over the internal knowledge base — embeddings pipeline, semantic retrieval, prompt orchestration, and evaluation — giving employees grounded answers over company documentation.
• •Led migration initiatives modernizing legacy systems on Harvard's platform while maintaining 99.9% uptime and seamless cross-service data synchronization.
• •Reduced API response times by 35% through performance optimization and caching strategy improvements.
• •Automated workflows with CRON jobs on the Behavox engagement, increasing team efficiency by 30%.