What AI products has Pablo Italo shipped?

Iluy, a controlled AI assistant with multi-provider LLM routing and content standards, used by 1,500 early users on web, SMS and WhatsApp; and an AI-assisted voice-signature workflow in AorForms, an e-signature SaaS with 20,000+ users.

Which LLM providers and tools does he work with?

OpenAI, Anthropic, Google and xAI models, with LangChain and the Vercel AI SDK, inside Next.js, NestJS and Django applications.

Can you add AI features to an existing product?

Yes. Most AI work starts inside an existing product: a feature is scoped around one user task, shipped behind guardrails, and measured before it grows.

How are LLM cost and reliability handled?

Through routing between providers, caching, fallbacks, moderation and evaluation on real prompts, so each request uses a model that fits its cost and latency budget.

AI Product Engineering

Build

Design and delivery of LLM features and AI products that real users rely on: RAG, AI agents, multi-model routing, moderation and AI-assisted workflows, built around cost, latency and quality.

At Iluy.ai, Pablo Italo designed LLM routing across OpenAI, Anthropic, Google and xAI and shipped web, SMS and WhatsApp access to 1,500 early users. At AorForms he delivered an AI-assisted voice-signature workflow tied to verification.

What it includes

Scope

01
Use-case discovery: where AI actually helps the user's task.
02
LLM app architecture: RAG, agents, tool use and multi-provider routing.
03
Guardrails: moderation, prompt management, evaluation and fallbacks.
04
Cost and latency tuning across model providers.
05
Production delivery: APIs, UI, monitoring and iteration.

Best fit

Who it is for

01
Shipping an LLM feature or AI product to production
Proof: Iluy.ai, AorForms.com
02
Routing between model providers to balance cost, latency and quality
Proof: Iluy.ai
03
Turning complex data or documents into answers users can act on
Proof: Climate4Buildings, AorForms.com
04
Not a fit
Research-only work on training new foundation models.
05
Not a fit
Demos that are not meant to reach production users.

Case studies

Evidence

01
Controlled AI assistant platform
Iluy.ai
Built a controlled AI assistant with model routing, content filters, and multi-channel access.
Role: Senior Full-Stack & AI Engineer (Contract)
Read the Iluy.ai case study
02
Compliance and e-signature SaaS
AorForms.com
Led a compliance-focused e-signature SaaS from early build to production users across document workflows, templates, SMS/email sending, and agent operations.
Role: CTO & Tech Lead, hands-on (Contract)
Read the AorForms.com case study
03
ClimateTech and architecture
Climate4Buildings
Led architecture and full development for an early-stage climate and energy analysis platform.
Role: CTO & Tech Lead, hands-on (Contract)
Read the Climate4Buildings case study