AI engineering
Custom AI Software — RAG, LLM Apps & AI Agents
I build production AI software that ships: RAG systems over your data, LLM applications, AI agents and chatbots, and AI features embedded into your app or CRM — full-stack, from interface to model.
AI software that actually ships
Plenty of AI projects stall at the demo stage. I build the version that goes to production — with real data paths, error handling, guardrails, and a UI your users and your team can actually trust. The goal is a measurable outcome: faster support, fewer manual steps, better lead quality, a product feature that converts.
Because I am full-stack, one engineer owns the whole thing: the interface (Vue/Nuxt), the backend and APIs (Laravel, Python/FastAPI), the integrations, and the AI layer on top. You do not have to coordinate three contractors to ship one workflow.
What I build
- AI agents and chatbots wired into your tools and workflows
- RAG systems that answer from your own documents and data, with sources
- LLM-powered automations: drafting, classification, extraction, routing
- Multi-step content pipelines with human approval gates, safety auditing, and per-run cost ceilings
- AI features embedded into existing products, dashboards, and CRMs
- AI-powered CRM automation (SuiteCRM, GoHighLevel) — reporting, lead handling, follow-up
- Provider-flexible architecture (OpenAI, Anthropic, open-weight) to avoid lock-in
Proof from real AI builds
For a marketing client I built a production AI content pipeline (FastAPI, Vue 3, PostgreSQL, LangChain) that turns campaign-performance data into expert-reviewed, SEO-optimized articles through a multi-step LLM workflow — with human approval gates, subject-matter safety auditing, a provider-agnostic engine (Anthropic/Gemini), per-run cost ceilings, and live keyword metrics. It is a real example of agents and RAG running safely at scale.
Other recent AI work includes an AI voice-agent calling-bot management portal built on SuiteCRM — campaign management where each campaign runs multiple servers with restart and configuration controls, and AI voice agents with per-agent voice generation, call scripts, and noise profiles. I also built a Vue front end with a Python FastAPI backend for a real-time call-tracking and analytics platform showing live campaign and call-flow dashboards. Two AI-powered SuiteCRM add-ons (an AI Dashlet Generator and a Business Card Reader) are published on the SuiteCRM Store.
Frequently asked questions
What kind of AI software do you build?
Custom AI applications that ship to production: AI agents and chatbots, retrieval-augmented generation (RAG) systems over your own documents and data, LLM-powered workflows and automations, and AI features embedded into existing products and CRMs. The focus is useful, reliable systems with real data paths and guardrails — not demos.
What is RAG and do I need it?
RAG (retrieval-augmented generation) grounds an LLM in your own content — docs, tickets, product data, a knowledge base — so answers are accurate and cite real sources instead of hallucinating. You need it whenever you want an AI assistant that answers from your data rather than generic internet knowledge.
Can you add AI to my existing app or CRM?
Yes. I build full-stack (Vue/Nuxt + Laravel/Python FastAPI) and integrate the AI layer into what you already run — including SuiteCRM and other CRMs. Recent work includes an AI voice-agent management portal built on SuiteCRM and a Vue + FastAPI real-time call-tracking and analytics platform.
How do we start, and how fast can I see something working?
Book a 30-minute build call or email the brief. We scope the highest-value workflow first and I build a working version with real data — typically a usable demo within about two weeks, then we harden it for production.
Which models and tools do you work with?
Provider-flexible: OpenAI, Anthropic, and open-weight models, with vector stores and orchestration chosen to fit your data, budget, and privacy needs. The architecture is built so you are not locked to one vendor.
Related
Tell me the CRM problem. I'll tell you what to build first.
Send the SuiteCRM migration, plugin idea, integration, or broken workflow. You'll get a clear build path and a working demo in two weeks — or you don't pay.