AI / LLM Engineer

I orchestrate AI agents into
systems that ship.

Agentic systems, LangGraph orchestration, and RAG architectures — designed, built, and deployed in production.

View my work Get in touch
About

Who I am

An AI/LLM engineer who turns manual processes into scalable agentic architectures.

I build the systems that make AI act — multi-agent orchestrations and retrieval pipelines that move from prototype to production. My focus is agentic design with LangChain and LangGraph, RAG over Postgres and pgvector, and the prompt engineering that makes all of it reliable enough to put in front of real users over channels like WhatsApp.

Underneath the AI work is a full-stack and data foundation — Python, TypeScript, Node, and React, plus the database and automation tooling to wire services together end-to-end. It means I can take an idea from architecture through deployment myself, rather than stopping at the demo.

Sergio Cuéllar Almagro — portrait
Expertise

Capabilities

Six areas where I build — flagship tools highlighted.

  • Agent Orchestration

    • LangGraph
    • LangChain
    • Multi-agent routing
    • Tool use
    • Stateful agents (memory)
    • Human-in-the-loop

  • Retrieval / RAG

    • RAG
    • RAFT
    • pgvector / HNSW
    • Embeddings
    • Semantic search

  • LLM in Production

    • Prompt engineering
    • Evals (LLM-as-judge)
    • Observability (Langfuse)
    • SSE streaming

  • Machine Learning & Data Science

    • scikit-learn
    • PyTorch / Keras
    • Deep learning (CNNs)
    • Computer vision (OpenCV)
    • NLP (NLTK)
    • Grad-CAM (explainability)
    • Pandas

  • Engineering

    • Python
    • FastAPI
    • TypeScript / Node.js
    • React / Next.js
    • PostgreSQL / Supabase
    • Docker
    • Hexagonal architecture

  • Automation & Integration

    • N8N
    • REST API integrations
    • WhatsApp / Meta API
    • ZohoDesk API
    • CI/CD
    • ETL pipelines
AI Showcase

How I build agentic systems

The real architectures behind my work — agents are stars, orchestration is the lines between them.

The orchestration and RAG diagrams are the real architecture behind Kynetix — my full-stack AI training & nutrition coach. The guardrail is the core of Agentic Job Engine, the job-search engine I built and use in my own search.

Multi-agent orchestration

A LangGraph supervisor classifies each message and routes it to one of six specialists — single-workout or full-week programming, profile updates, the nutrition sub-graph, post-workout analysis, or general chat. Programming grounds plans in a pgvector RAG store and deterministic services (1RM, fatigue, TDEE); analysis reads the relational workout history; nutrition commits via a human-in-the-loop confirm. Conversation state persists through a Postgres checkpointer and replies stream token-by-token over SSE.

Guardrails: anchored generation

The model never writes free prose about a candidate's experience — it drafts a CV out of source_key references to Profile items. A pure validator, with no database, network or LLM in the loop, rejects any key the Profile doesn't contain, so an invented job or skill fails before anything is saved. Only a draft that passes has its keys resolved back to real Profile text and rendered to PDF for a human to review.

RAG retrieval pipeline

Reference knowledge (kinesiology, nutrition) is embedded and stored once; at query time the question is embedded, matched against a pgvector HNSW index, and the top-k chunks ground the model's answer — no ungrounded guessing.
Ask my portfolio

Ask my portfolio

A live RAG assistant, grounded in my real experience — the architecture above, working.

Try asking

Prefer email? info@sergiocuellar.dev
Projects

Flagship work

Selected agentic and LLM systems I designed and built end to end.

Agentic Job Engine — Grounded Job Search & CV Tailoring

A local-first agentic job search: LangGraph pipelines discover and score offers against a structured Profile, then draft tailored CVs whose every line cites a Profile key — so a deterministic validator rejects anything the model invented.

  • LangGraph
  • FastAPI
  • OpenAI
  • SQLite + sqlite-vec
  • APScheduler
  • Playwright
  • React PWA
Case study Code

Kynetix — AI Gym Trainer & Nutrition Coach

A full-stack AI training and nutrition coach: a FastAPI + LangGraph backend running a streaming, multi-node agent grounded by RAG over pgvector, with a React PWA client.

  • LangGraph
  • LangChain
  • FastAPI
  • OpenAI
  • PostgreSQL + pgvector
  • Redis
  • React PWA
Case study Private repo

Ask My Portfolio — RAG Chat API

The grounded RAG chat that answers questions about me on this very site — a single Cloudflare Worker with SSE token streaming, Supabase pgvector retrieval, and Turnstile + KV abuse guards.

  • TypeScript
  • Cloudflare Workers
  • OpenAI
  • Supabase + pgvector
  • Turnstile
  • Workers KV
  • SSE
Case study Code Live
More of my work on GitHub
Experience

Where I've shipped

A short spine of AI-first roles, most recent first.

  1. Simply Trade IN

    AI Automation Developer

    Nov 2025 – Mar 2026

    • Built RAG support agents that encode domain expertise into grounded, accurate answers.
    • Implemented semantic retrieval on Supabase with Postgres + pgvector.
    • Shipped a conversational sales engine and automated social content through N8N pipelines.
    • Automated operational pipelines across several departments with N8N, weaving AI into day-to-day workflows.
    • Built Python and N8N ETL dataflows integrating multiple third-party APIs, backed by relational Postgres/Supabase schemas.

  2. BoosterPrompt

    Web App Developer — Backend / Agentic Systems

    Jun 2025 – Mar 2026

    • Built a multi-agent conversational booking system with LangChain and LangGraph, orchestrating tool-use across specialized agents.
    • Delivered end-to-end WhatsApp booking through the Meta API, taking the system to real users in production.
    • Hardened agent reliability with systematic prompt engineering.
    • Engineered the backend and data flows in Node.js, covered by unit and integration tests in Vitest.
    • Built a notifications service on the Resend API and designed the corporate website.

  3. Freelance

    Web Developer & Data Analyst

    Mar 2023 – Nov 2025

    • Designed and shipped web applications for clients, with SEO/SEM optimization.
    • Ran business data analysis — sentiment and KPIs — in Python (Pandas, NLTK).
Contact

Let's build something

Open to AI/LLM engineering roles. The fastest way to reach me is email.

Email me Download résumé (PDF)
GitHub LinkedIn

Based in Marbella, Spain