Hi, I am Kristijan Soldo.
I'm a software developer and AI architect based near Zagreb, Croatia. Over a decade across B2B SaaS on PHP/Symfony/Laravel, Python, Vue, and Kubernetes-based DevOps, now focused on embedding production-grade AI into mature systems without blowing them up.
Selected Engineering Works
Provirium: AI-Guided Automotive Diagnostics Platform
A diagnostics platform for mechanics and workshops that turns symptoms, fault codes, and live vehicle telemetry into provable findings, so parts get replaced based on evidence, not guesswork. Mobile app connects to the vehicle via an OBD adapter; the web dashboard tracks every session, graph, and confirmed repair.
Legend Energy: AI Layer for Multi-Tenant SaaS
Designed and built the AI service layer for an industrial energy analytics platform serving large US facilities. Semantic query planning, LLM orchestration, strict multi-tenant data isolation, and async processing over NATS/JetStream, integrated with an existing Laravel/Python/Vue stack on Kubernetes.
Autonomous AI Operations Agent: LangGraph, a multi-model LLM stack, and human-in-the-loop learning
A production AI agent that autonomously triages and processes customer-support tickets for an NDA-protected SaaS. It proposes every response to a human reviewer in the CRM, executes only what's approved, and learns durable rules from every answer. Replaces a ~2,000-line n8n monolith with a modular Python and LangGraph service running a multi-model LLM stack (DeepSeek for reasoning, Qwen3-VL for vision, Whisper for audio) to keep inference cost predictable at scale.
The Soldo Journal
The Anti-AI Design Philosophy: Finding Authenticity in a Monochrome Web
Why standard SaaS layouts and purple gradients look the same, and why strong contrast, intentional typography, and brutalist accents feel so refreshing today.
A Decade of Frontline Engineering: 3 Rules I Write Code By
Over 10 years in full-stack engineering teaches you that tools fade but principles remain. Here are the rules that keep software stable, clean, and fast.
Achieving Sub-100ms Interaction Times on Modern Web Applications
Practical front-end performance techniques that move beyond lighthouse scores into real user perceptual responsiveness.
Initiate Connection
You can reach out to me via
Message sent
Thanks for reaching out. I'll get back to you within a day or two.