Tech lead, full stack and agentic AI. Remote from El Salvador for US teams.
I put LLMs into products that people actually use — RAG pipelines, scoring and classification workflows, Claude-powered features — and own the Django and Next.js they sit in, plus the CI/CD and AWS underneath. Seven years shipping product, three leading the engineers who ship it.
$ whoami
david — tech lead · full stack & agentic ai
$ pwd
epallet/ — react + django, leading the engineering team
$ cat now
teaching a product to read, and a team to build with agents
$ open davidquinta.techNot a list of everything — four I would happily walk you through.
jobfit — where deterministic rules end and a model should start. Ingests 846 job postings, cuts them to 135 for free, then spends one structured LLM call each on what survives. The README is blunt about what is measured and what is not, which is most of the point.
DefaultClassifier-RestAPI — operating a model rather than deploying one. Registry, versioned status history, every prediction logged, and A/B tests scored against outcomes that only arrive months later.
bifrost-data-acquisition — a FastAPI ingestion backend: routers over a shared model layer.
rag-evaluation-test — short enough to read in one sitting. Retrieval scoring with FAISS, and why a cosine score is a smoke test rather than an evaluation.
- Agentic AI — Claude API, MCP, agent tooling, evals
- Backend — Python, Django, FastAPI, PostgreSQL
- Frontend — TypeScript, Next.js, React
- Infra — GitHub Actions, Docker, AWS
MSc in AI & Big Data. I work in English and Spanish, and I organize GDG San Salvador.
davidquinta.tech · LinkedIn · hello@davidquinta.tech
Open to senior and lead roles — contract, freelance or full-time.
