AI ENGINEER · THESSALONIKI, GR

IOANNIS PEGIADIS.

I design and ship production LLM systems — RAG pipelines, AI agents, and the infrastructure underneath. Six years. Enterprise clients. Software that runs.

0years shipping
0production apps
0MSc degrees

SELECTED WORK

01/05

systems in production

WORKFLOW · LIVE React UI SSE streaming FastAPI RAG router · re-rank Pinecone cited retrieval GPT-4 / Claude multi-model MySQL usage tracking DUAL-STREAM RAG — DOCUMENTS (CITED) + GENERAL LLM, TOKENS STREAM BACK

Future Cats — Enterprise RAG

FUTURE CATS · CONSULTING · 2025—NOW

Natural-language search across compliance docs and label regulations for an enterprise client in the food industry. Hybrid retrieval + re-ranking, OpenAI & Claude orchestration, OCR ingestion, per-tenant permissions, streaming responses. Label-compliance agents cut manual review from hours to minutes.

FastAPI / Pinecone / OCR / WebSockets

Vetly clinic dashboard
WORKFLOW · LIVE Owners + vets · JWT Caddy vetly.gr Next.js 15 SSR FastAPI SQLAlchemy Postgres multi-tenant Redis cache Llama-3 local LLM Resend email 100% PRIVATE LLM — ZERO PER-TOKEN COST · SELF-HOSTED ON HETZNER

Vetly — Veterinary platform

FOUNDER · SOLO BUILD · 2025—NOW

Multi-tenant clinic management: records, scheduling, billing, pet-owner app. Local Llama-3 summarizes patient history — full privacy, zero token costs. Self-hosted on Hetzner: Docker, Coolify CI/CD, Grafana observability.

Next.js / FastAPI / PostgreSQL / Local LLMs

WORKFLOW · LIVE Copernicus Sentinel-2 raster Earth Engine NDVI · NDWI Weather APIs temp · precip Stations field sensors Python parallel workers AWS S3 raster archive PostGIS sub-200ms queries ML models training data Slack alerts IDEMPOTENT · RESTARTABLE · 400—800× FASTER THAN NAIVE BASELINE

Spatial data at scale

ECO-DEVELOPMENT · CONSULTING · 2024—NOW

PostGIS handling millions of spatial records, tuned for complex joins. Satellite pipelines from Copernicus & Google Earth Engine into S3 with automated ingestion and Slack alerting. Legacy stack migrated to .NET / React.

.NET / PostGIS / Earth Engine / CI-CD

WORKFLOW · LIVE Spec plan as written Claude Code skills · MCP · agents Review human-in-loop Ship prod SPEC-DRIVEN SDLC — REJECTED WORK LOOPS BACK TO AGENTS

Agentic dev toolkit

PERSONAL FRAMEWORK · 2025—NOW

My operating system on Claude Code: custom skills, subagents, MCP servers, persistent memory. Spec-driven SDLC — plan as spec, delegate to agents, review human-in-the-loop. One engineer, team-scale output.

Claude Code / MCP / Multi-agent

Sportsholics homepage
WORKFLOW · LIVE Visitors readers Caddy TLS Next.js 15 SSR + ISR · SEO Strapi 5 headless CMS Postgres 1780 articles 1780 ARTICLES — FOOTBALL · BASKETBALL · F1

Sportsholics

CONTENT PLATFORM · 2024

High-traffic Greek sports analytics platform on a custom headless CMS. Performance engineered, not hoped for.

Strapi / Next.js / PostgreSQL

ABOUT / METHOD

I work AI-first: every feature starts as a written spec, implementation is delegated to agentic coding systems, and nothing ships without human review. That's how one engineer delivers like a team — without the slop.

Stack

Python · TypeScript · FastAPI · Next.js · PostgreSQL · pgvector · Pinecone · Docker · Grafana

AI / LLM

RAG · hybrid retrieval · re-ranking · agents · MCP · OpenAI · Anthropic · local Llama-3

Background

MSc Data Science & AI (York) · MSc Business Analytics (Sheffield) · based in Greece, working remote-first

EXPERIENCE

Track record

2025—NOW AI Engineer (Consulting) Future Cats Production RAG & agent systems
2024—NOW Senior Software Engineer (Consulting) Eco-development Spatial data & backend systems
2023—2024 Software Engineer BetAdvanced High-frequency betting APIs
2022—2023 Data Engineer Eco-development ETL & ML data pipelines
2022—2024 MSc Data Science & AI — Distinction-level University of York York, UK
2020—2022 MSc Business Analytics — Merit University of Sheffield Sheffield, UK

CAPABILITIES

Skills

AI & LLM

RAG pipelinesHybrid retrievalRe-ranking AI agentsMCP serversPrompt engineering OpenAI APIAnthropic APILocal LLMs (Llama-3) OCR ingestionLangChainPyTorch

Languages

PythonTypeScriptJavaScript C#SQLBash

Backend & Frontend

FastAPINext.jsReact .NET CoreWebSocketsREST APIs

Databases

PostgreSQLpgvectorPostGIS RedisPineconeMySQL

AI-assisted development

Claude CodeCustom skillsSubagents Spec-driven SDLCWorkflow orchestrationHuman-in-the-loop

DevOps & Observability

DockerGitHub ActionsCoolify HetznerGrafanaPrometheusAirflow

GET IN TOUCH

LET'S BUILD
SOMETHING REAL

ioannispegiadis97@gmail.com