AI Agents & RAG Systems
Assistants and document intelligence grounded in your data — with verifiable citations, evaluation and guardrails. From corpus ingestion and vector search to a deployed, authenticated product.
System Online — La Paz, Bolivia · GMT-4
AI ENGINEER
I build AI products that actually ship — retrieval-augmented generation, LLM agents and automation — backed by solid full-stack engineering in .NET, React, Python and FastAPI. Currently operating a production legal-AI platform used by professionals in Bolivia.
01 // About
I'm Diego Humerez, an AI and full-stack software engineer based in La Paz, Bolivia. I design and ship AI-powered products — retrieval-augmented generation (RAG) systems, LLM agents and document intelligence — together with the full-stack engineering they need to reach production: .NET, React, TypeScript, Python and FastAPI.
My current flagship is Asesor Legal IA, a production legal-AI platform for Bolivian law: a grounded RAG pipeline over 9 legal codes and 4,170+ articles, where every answer carries citations you can verify against the official source — no hallucinated law.
Before going deep on AI I spent years shipping enterprise software, working remotely with US companies and building SaaS for businesses across Latin America. That mix means I don't just prototype models — I deliver authenticated, containerized, monitored products running on real infrastructure.
Off duty you'll find me building Gunpla, queuing Dota 2, hunting street food or planning the next trip.
02 // Mission Types
I help businesses in Bolivia, Latin America, the US and Europe ship custom software: web applications, business systems, process automation and AI-first products — from the first scoping call to a deployed, monitored product, in English or Spanish.
Assistants and document intelligence grounded in your data — with verifiable citations, evaluation and guardrails. From corpus ingestion and vector search to a deployed, authenticated product.
Multi-tenant SaaS, real-time dashboards and ops tools with enterprise-grade backends and polished frontends. Payments, invoicing and tax integrations (incl. SIN Bolivia) handled.
Dockerized deployments on your VPS or cloud, CI/CD pipelines, reverse proxies and TLS, plus scraping and data pipelines that keep your systems fed automatically.
03 // Arsenal
AI & Data
Backend
Frontend
DevOps
04 // Deployments
AI legal advisor for Bolivian law, in production with real users. Grounded RAG over 9 legal codes; multiple analysis modes (case strategy, legal opinions, bill drafting); per-customer authentication and isolated workspaces.
4,170+ articles indexed — every citation links to the official text. Verifiable, not hallucinated.
Open PlatformEnd-to-end restaurant operations — live table map, order tracking, kitchen display and digital receipts, with SIN tax-compliance integration built for the Bolivian market.
Runs the full dining-room → kitchen loop in real time.
View ProjectDebt ledger for individuals and small businesses with digital receipt storage: who owes what, receipts attached as proof, pending vs settled — one clear picture.
Paper receipts → searchable, shareable records.
View ProjectBuilding administration platform for LATAM condominiums — expense tracking, unit management, common-area bookings and resident communication for property admins.
Next deployment in the pipeline.
More in the lab — insurance quoting, WhatsApp legal assistant, internal tooling → github.com/dhumerez
05 // Off-Duty
06 // Intel
Custom web applications, business systems (point-of-sale, inventory, admin platforms), process automation, and AI-first products: assistants that answer from your own documents (RAG), LLM-powered workflows and chatbots, plus integrations between the tools you already use.
Yes — I'm based in La Paz and build for businesses across Bolivia and Latin America in Spanish, and for US and European teams in English. Everything is remote-friendly (GMT-4), from the first scoping call to production support. I've already shipped products with Bolivian specifics, like SIN tax-compliance integration.
Usually, yes. Typical starting points: a chat assistant grounded in your documents or database, automatic document processing (OCR + data extraction), or automating a repetitive workflow with an LLM in the loop — integrated into the system you already run, with verifiable outputs instead of hallucinations.
A short scoping call (free) → a fixed-scope proposal with milestones, or an hourly arrangement for ongoing work. Small automations take days; full platforms take months. Send me what you need and you'll have an answer within 24 hours.