{
}
ERIKASIO

Production AI systems that generate revenue, not demos

I build LLM agents and RAG architectures that ship to production and deliver measurable ROI. 8 years of experience building ML systems for companies from $50M startups to $3.7B enterprises.

Book a Discovery Call
LangChain LangGraph OpenAI Anthropic Claude Code Google Gemini AWS GCP Weaviate Pinecone FastAPI
8 years building ML systems | Former engineer at CarVertical and HomeToGo
Case Studies

Systems in production

Global payroll and workforce management platform with 1000+ enterprise customers
$0
raised by client

$3.7B HR SaaS Platform

Production RAG system for customer support — handles conversational queries across HR compliance documentation spanning employment laws, payroll regulations, and country-specific requirements

Technical Highlights
  • LangGraph state machine orchestration
  • Hybrid search: BM25 + vector + RRF
  • Query classification routing
  • WebSocket streaming (1000+ concurrent)
  • LangSmith production monitoring
Result

Resolved critical architectural bottlenecks blocking launch. System now serves enterprise customers across 60+ countries.

European customs law SaaS company needing AI-powered compliance assistance
$0
per user as a SaaS feature

Legal SaaS — Regulatory Compliance RAG Agent

RAG agent for customs law compliance — calculates fees and tariffs for imported/exported goods, answers regulatory questions, grounds all answers in European customs law documents

Technical Highlights
  • Hundreds of legal docs, thousands of pages
  • Citation grounding with source attribution
  • Abstention when info unavailable
  • Commercialized as SaaS feature
Result

AI agent now sold as a revenue-generating product feature. Helped client win a government contract.

€168M revenue manufacturer with 1000+ employees, products sold in 60+ countries
0%
error reduction

Enterprise Furniture Manufacturer

LLMOps foundation and AI-driven ticket handling system for customer support automation

Technical Highlights
  • Automated ticket classification and routing
  • Response generation with quality controls
  • Monitoring and evaluation pipeline
  • Error catching before customer reach
Result

Reduced customer support automation errors by 40% through systematic LLMOps practices.

Background

8 years in production ML systems

CarVertical
ML Engineer
1.5 years
$50M vehicle history platform operating in 28 countries with 900+ data sources
What I Built
  • Information retrieval system matching scraped vehicle data across multiple sources
  • Entity resolution: matching "Volkswagen Golf 2005" to standardized versions
  • TF-IDF, entity matching, hybrid AI data mapping
86% cost reduction, 128x faster inference
HomeToGo
ML Engineer
5 years
Vacation rental marketplace that went public at €1.2B valuation, $300M+ annual revenue
What I Built
  • Learning-to-rank systems optimizing search result ranking
  • Search personalization and booking prediction
  • ML systems matching user intent to property listings
+9% increase in direct bookings through AI-powered search re-ranking
Process

How we work together

01

Discovery Call

Free, no obligation
  • Discuss your AI goals and current pain points
  • Identify where you're losing time or money
  • Assess if a production AI system makes sense
  • Get clarity on the path forward
02

Architecture & Scope

Technical blueprint
  • Map out data sources and retrieval strategy
  • Design orchestration and deployment architecture
  • Clear scope document with deliverables
  • Align on what "done" looks like
03

Build & Ship

Production-grade delivery
  • Build with production-grade infrastructure
  • Evals, monitoring, and observability from day one
  • System that runs reliably and can scale
  • Not a prototype — a production system
Services

What I build

RAG Systems

Retrieval-augmented generation for customer support, internal knowledge, compliance documentation. Hybrid search (BM25 + vector + reranking), citation grounding, abstention when uncertain. Built for accuracy, not just answers.

LLM Agent Systems

Multi-step AI agents orchestrated with LangGraph. Query classification, specialized workflows, structured outputs. Production-ready with error handling that keeps edge cases from crashing your customer experience.

Information Retrieval & Search

Search ranking, entity resolution, vector databases (Weaviate, Pinecone, OpenSearch). Learning-to-rank systems. Make your search results actually relevant.

Production AI Infrastructure

LLMOps foundations: evals, monitoring with LangSmith, streaming APIs (FastAPI + WebSockets), scalable deployment on AWS/GCP. Systems that don't break at 2am.

Ready to ship production AI?

Skip the agencies that build demos. Let's build a system that runs in production and delivers measurable results.

Book a Discovery Call