I Build AI Systems That Actually Work in Production_
20+ years shipping software. Now architecting AI agents, RAG pipelines, and multi-agent systems that scale, stay safe, and deliver ROI—not demos.
Born in Germany 🇩🇪 · Building from Batumi, Georgia 🇬🇪 · Working globally
Most AI projects fail.
Not because of the technology—but because teams treat AI like it's magic instead of engineering.
There's this belief that you can just prompt your way to a working product. I've seen it many times: teams chase demos that wow in meetings but break in production. They skip evaluation, skip architecture, skip safety—and then wonder why their AI agent goes rogue or costs $50K/month in API calls.
Why do I tell you this? Because I've learned that AI without discipline is just expensive guessing. It sounds simple, but it isn't. Production AI needs the same rigor as any other system—architecture, testing, measurement, iteration.
That's what I do. I build AI systems that work.
About
I'm Kevin—a builder who's been writing code and shipping products for over 20 years.
I started as a developer, became a founder, scaled teams from 2 to 50+ people, and led engineering orgs across startups, SaaS companies, and enterprise platforms. I've built marketplaces, e-commerce systems, insurer platforms, and infrastructure that served millions.
I got into AI in 2021 because I saw a gap that fascinated me: everyone was playing with demos, but almost no one was building systems that worked reliably at scale. AI required everything I'd learned about architecture, product thinking, and team leadership to actually work in production.
When I look back, this was one of the most relevant learnings of my career. The technology is exciting, but what makes it work is the discipline around it—the same discipline that makes any complex system work.
Since then, I've designed multi-agent systems, RAG pipelines, evaluation frameworks, and orchestration architectures that run 24/7, stay safe, and deliver measurable business outcomes.
I don't just know AI. I know how to build products, design architectures, lead teams, and ship software that doesn't break.
Currently
Founder
Oikei LLC
Previously
CTPO
kitchX
CTO
claimsforce
CTO
HAPPYCAR
Founder
Yopegu
Based in
Batumi, Georgia 🇬🇪 (Born in Germany, building globally)
I don't build AI toys. I build systems that ship.
My work sits at the intersection of architecture, engineering, and product strategy. I help teams move from "can we do this?" to "here's how we do this reliably, safely, and at scale."
Architecture & System Design
- →End-to-end AI system architecture (agents, RAG, orchestration)
- →Multi-agent workflows with LangGraph + supervisor patterns
- →Integration into existing stacks (web, mobile, WhatsApp, voice, APIs)
- →Cloud-native design for high availability and low latency
Engineering & Implementation
- →Structured reasoning pipelines & safe tool usage
- →Retrieval systems & vector search optimization
- →Event-driven & service-oriented architectures
- →Production-grade code that scales
Quality & Measurement
- →Evaluation frameworks (not vibes-based AI)
- →Guardrails & regression testing
- →KPI-driven architecture decisions
- →Continuous improvement loops
Strategy & Execution
- →High-stakes problem solving
- →Product-AI alignment
- →Team enablement & technical leadership
- →Fast iteration cycles based on real data
My Approach
AI needs discipline, not hype. I believe most AI projects fail because teams chase magic instead of building systems. I build systems.
Predictability > Vibes
If you can't measure it, you can't improve it. That's also why every AI system I build has evaluation baked in from day one.
Architecture > Prompts
Prompts are UI. Architecture is the foundation. I design systems that work because of their structure, not because of lucky prompt engineering.
Safety by Design
Guardrails, evals, regression tests—built in from the start, not bolted on when something breaks. I learned this the hard way.
Ship, Measure, Iterate
Fast feedback loops beat perfect plans. I ship early, measure everything, and iterate relentlessly. It's a process, not a one-time effort.
ROI or GTFO
AI should make money or save time. Everything else is a science project. I optimize for business outcomes, not technical elegance.
Systems > Magic
I don't believe in AI magic. I believe in well-designed systems that happen to use LLMs as one component.
Expertise
AI Engineering & Architecture
- •Multi-agent orchestration (LangGraph, supervisor patterns, state machines)
- •RAG systems (retrieval, indexing, chunking strategies, vector search)
- •Structured reasoning & chain-of-thought pipelines
- •Safe tool usage & function calling
- •Evaluation frameworks (regression testing, quality scoring, A/B testing)
- •Guardrails & safety layers
- •Prompt engineering at scale
Software Architecture & Systems
- •Domain-Driven Design (DDD)
- •Clean Architecture & modular systems
- •Event-driven architecture (EDA)
- •Service-oriented architecture (SOA)
- •Cloud architecture (GCP, AWS)
- •High-availability / low-latency design
- •API design & integration patterns
Product & Team Leadership
- •AI product strategy & roadmapping
- •KPI frameworks & operational measurement
- •Cross-functional team leadership
- •Continuous delivery & experimentation culture
- •PMF discovery & iteration
- •Technical hiring & team scaling
Tech Stack
Experience
Founder
Oikei LLC
Engineering Rational AI Systems. We build AI applications with clarity, safety, and purpose.
- →Focusing on production-grade multi-agent systems and AI applications for business and society
- →Building retrieval-based AI with high factual accuracy and evaluation frameworks
- →Ensuring safety, PII protection, and policy-aligned behavior
- →Designing cloud-native architectures that scale cleanly
Impact: Rational, safe, and human-centered AI systems built from Batumi, Georgia.
Freelance AI Architect
Qualimero
Building production AI systems for enterprise workflows.
- →Designed and deployed multi-agent systems handling complex business logic across support, sales, and operations
- →Built evaluation pipelines that measure quality, catch regressions, and enable continuous improvement
- →Architected guardrail systems ensuring safe, predictable AI behavior in production
- →Created orchestration frameworks for reliable agent coordination and state management
Impact: Production systems running 24/7 with measurable quality metrics and zero critical incidents.
Chief Technology & Product Officer (CTPO)
kitchX
Led product and tech strategy for AI-driven e-commerce kitchen platform.
- →Built WhatsApp AI agent that increased lead qualification from 6% → 70%
- →Designed AI-first customer journey from discovery to purchase
- →Integrated AI across web, mobile, and messaging platforms
Impact: 10x improvement in conversion, >$500K saved in human ops costs annually.
CTO & Product Lead
claimsforce
Led 30+ person product, tech, and data organization for insurer SaaS platform.
- →Built cloud-native SaaS platform from ground up
- →Designed scalable architecture supporting multiple enterprise clients
- →Led product strategy and tech roadmap
- →Supported $8.4M Series A with robust technical foundation
Impact: Platform serving multiple insurers, processing thousands of claims, enabling successful Series A.
CTO
HAPPYCAR
Built and scaled travel marketplace and engineering organization.
- →Scaled engineering team from 2 → 20
- →Designed and built marketplace platform and cloud infrastructure
- →Established technical culture and engineering practices
- →Led hiring, team structure, and technical strategy
Impact: Platform serving hundreds of thousands of users across European markets.
Earlier Roles
Various
Foundation years building software expertise.
- →Founder, Yopegu – Built and sold B2B SaaS product
- →Mobile & Web Engineer – Early smartphone era development
- →Team Lead – Engineering management and product development
What I've Built
Concrete proof points from production systems
AI Sales & Support Agent (WhatsApp)
Qualified 70% of inbound leads automatically—at scale, 24/7, zero human handoff.
RAG-Based Product Advisor
Built retrieval system with 95%+ factual accuracy, serving thousands of queries daily.
Multi-Agent Orchestration System
Production system coordinating multiple specialized agents across support, sales, and ops workflows.
Evaluation & Quality Pipeline
Framework that catches regressions before customers do. Automated testing, quality scoring, and continuous monitoring.
Enterprise SaaS Platform (Insurer Tech)
Cloud-native platform serving multiple enterprise clients, processing thousands of claims.
Travel Marketplace (HAPPYCAR)
Built platform from ground up, scaled to hundreds of thousands of users across European markets.
Who I Work With
I work with teams who are serious about AI.
✓You're a great fit if
- →You need AI systems that work in production, not just demos
- →You're building products where reliability and safety actually matter
- →You have real users, real data, and real business metrics
- →You want to move fast but build things properly
- →You're willing to invest in architecture, not just prompts
✗You're not a fit if
- →You want someone to "just make ChatGPT for X"
- →You're looking for the cheapest option
- →You expect AI to magically solve problems without engineering
- →You're not willing to measure and iterate
Types of Projects
Let's Connect
I take on a limited number of projects each year.
If you're building AI systems that need to work in production—not just look good in a deck—let's talk.
I work with founders, CTOs, and product leaders who are solving hard problems and need an experienced architect to help them ship reliable AI systems.
I typically respond within 24-48 hours.