Portfolio · 2026

Vasile Bogdan Godja

Applied AI Engineer · AI Systems & Workflow Automation

Founder-operator turned applied AI engineer. I build production AI systems — multi-agent orchestration, RAG, and workflow automation — for problems I've personally worked inside: EU funding, public-sector administration, and hospitality operations. 10+ years running businesses end-to-end, including €500K+ in EU-funded projects. I think like an operator and ship like an engineer.

Selected projects

Three production AI systems I designed, built, and own end-to-end — each one a deep case study in applied LLM systems, multi-agent orchestration, and the operational reality of a specific vertical.

GovTech · Tax SaaS

PrimărIA — multi-tenant tax SaaS for Romanian municipalities

SaaS-first, AI-ready. Multi-tenant via Postgres RLS. ANAF-compliant.

Production-grade tax administration platform for ~2,800 Romanian communes — deterministic tax engine for buildings/land/vehicles, PatrimVen XML export to ANAF, dual-portal architecture, AES-256-GCM CNP encryption, PWA offline for rural areas. The AI agent surface is designed-for and lands on top of a substrate that already answers correctly without it.

Next.js 14TypeScriptPostgreSQL + RLS Prisma 6PgBouncerRedis + BullMQ MinIONextAuth v5TOTP 2FA @react-pdf/renderernext-intl (RO/HU/EN)Docker
Vertical AI · Hospitality

MitchAI — vertical operational AI for hospitality

Multi-agent ops suite on hybrid local + cloud LLM infra.

Ops, Media, Menu, Legal, and Stock agents coordinated by a planner agent, wired together with n8n. WhatsApp-first approval surface; local LLMs for high-volume tasks so per-outlet economics stay viable for independent operators. Grounded in 10+ years on the operator side of the problem.

PythonFastAPIn8n PostgreSQLRedisDocker Ollama (local)ClaudeOpenAI WhatsApp API

More from the workshop

Supporting infrastructure, vertical experiments, and client work — each one a different angle on the same operator-engineer profile.

Infra · Multi-provider LLM

AI Gateway — unified multi-provider LLM middleware

OpenAI-compatible router for OpenAI, Anthropic, Gemini, xAI, Groq, and Perplexity.

A single secure entry point for every LLM call across my projects. Token auth, security headers, CORS, rate limiting, structured logging, and Cloud Run-optimised cold starts. The boring infrastructure layer that makes everything else above it look easy.

TypeScriptNode.jsHelmetPino DockerGoogle Cloud Run
Vertical AI · MCP

Hospitality Marketing Assistant — MCP-server marketing stack

Talk to your marketing stack in natural language, via Claude.

Modular Python services exposed over Model Context Protocol — brand voice, content automation, social analytics — orchestrated for hospitality operators. Companion / precursor to the MitchAI Media agent, and a working example of MCP-first product design.

PythonMCPPostgreSQLDocker Claude Desktop
Client work · Swiss mobility

Lugano Car Rental — premium mobility platform for Ticino

Full-stack booking platform + bilingual business plan.

Frontend, backend, Docker deployment, and a full business plan (Romanian + Italian) for a Swiss car-rental operator. Shows the operator-engineer pattern end-to-end — product design, implementation, and the commercial framing that turns it into a real business.

HTMLTypeScriptPythonDocker
Experimental · Cosmos SDK

Mitch Chain — Cosmos SDK L1 for the Mitch ecosystem

MTC coin · loyalty + AI credits · IBC-ready.

Application-specific Layer-1 designed to settle loyalty points across Mitch outlets and meter AI-credit usage inside MitchAI. Cosmos SDK + Ignite scaffold, CometBFT consensus, planned AWS deployment with Terraform + systemd + Prometheus. Research-grade work-in-progress — published for visibility, not yet a mainnet.

GoCosmos SDKIgnite CLICometBFT ProtobufTerraform (planned)

About

I bridge two worlds most candidates don't: 10+ years of hands-on business operations and 6+ months of full-stack applied AI engineering.

Before EuFund, PrimarIA, and MitchAI, I spent a decade running high-performing teams in the UK hospitality industry — Restaurant Manager at 108 Brasserie (Marylebone Hotel), Hospitality Manager at London Business School (50-person team serving 1,000–2,000 guests/day), General Manager at The Crown (Mosaic Pub & Dining), Assistant GM at Head of the River (Fuller's, Oxford).

I'm also the founder of Mitch from Transylvania, a London food-market business, and former GM of a Romanian restoration company specialising in UNESCO and heritage-listed sites — where I personally secured and managed €500K+ in EU grant funding end-to-end.

That operator background is the design substrate for everything I build. I'm opinionated about which 20% of a workflow not to automate, and honest about which AI surfaces deserve confidence and which deserve a hard "refuse and route."

Languages: English (fluent), Romanian (native), Spanish & Italian (intermediate), French (beginner).

Skills

Hands-on across the stack — from agent orchestration down to the Postgres indexes that make retrieval not embarrassing.

AI engineeringLLM apps · multi-agent orchestration · RAG / vector search · structured outputs · multi-model routing · streaming · evals · observability · cost optimisation
Backend & infraPython · FastAPI · Next.js 14 / TypeScript · PostgreSQL · pgvector · Qdrant · Drizzle · Redis · Docker · Google Cloud Run · REST · async I/O
Automation & workflown8n · workflow design · event-driven systems · webhook integration · hybrid local + cloud LLMs
ToolingClaude Code · Codex · Gemini CLI · Cursor · GitHub Actions · Cloud Build · gitleaks · Playwright · Vitest
Operations & deliveryProduct thinking · P&L ownership · stakeholder management · compliance workflows · EU-funded project delivery · cross-functional leadership
LanguagesEnglish (fluent) · Romanian (native) · Spanish (intermediate) · Italian (intermediate) · French (beginner)

Documents

Everything in one place — CV in two flavours, plus all three case studies.