Things built.
DATHHUB.
A summonable Jarvis for running life and work from one desktop HUD. Persistent workspace chats, Google Calendar, goals, project awareness, local voice, NeuroVault memory, and a permission broker that keeps every agent action visible and governed.
View case study
NeuroVault.
A Tauri desktop app that gives Claude persistent memory across conversations. Markdown vault, knowledge graph, hybrid retrieval — outperforms RAG at 96% hit@3, ~275 tokens/answer, and runs entirely on your machine.
View case studyAEGIS.
Maritime shipping risk platform aggregating 7+ sources across 4 database technologies. LLM-powered CrewAI agents for risk, route planning, and trade compliance — with W3C PROV lineage and hallucination checking.
View case study
London Synergy.
Pipeline tessellating Greater London into 55,000 H3 hexagons — LandScan rasters, ONS Census, OSM POIs enriched with graph-centrality. Spatially cross-validated XGBoost; Burt's Structural Hole Theory applied to retail siting.
View case studyRiskTerrain.
A 6-node LangGraph pipeline that ingests live geopolitical events (USGS, NewsAPI), traverses a 154-company supply-chain graph in SurrealDB, and scores cascading exposure in under 8 seconds. Hybrid graph-vector retrieval with sentence-transformer embeddings.
View case study
DS2 Workspace.
The private operating system for a two-founder digital-solutions consultancy. An agentic copilot with model routing and a confirm-gate, six owned AI micro-engines, and a full pipeline / leads / content workspace — one authenticated Next.js app, fully themed light and dark.
View case studyNodebook.
An on-demand knowledge compiler. Hand it a topic, a link or a question and it compiles a beautiful, cited book — chapters, math, diagrams — then wraps it in a spaced-recall loop so you actually keep it. Next.js 15 web plus an Expo iOS/Android app on Supabase.
View case studyOpen live ↗
DreamBug.
A personalized children's storybook app. Parents pick a hero; DreamBug assembles illustrated, age-appropriate books on-demand with a coherent cast and world on every page — from a local, cost-free art pipeline instead of a live paid image API.
View case studyOpen live ↗Index · 04 more
/ all open in case studies
London Housing & Crime
Spatio-temporal pipeline merging 1M+ records into a predictive Opportunity Index. XGBoost regressor tuned via Optuna, R² = 0.92.
DataScrub
Privacy-first browser-based data cleaning agent. Rule-based + LLM, optional local Ollama inference, no backend.
Dataportfolio.co.uk
Production SaaS that automates portfolio creation for data professionals — no-code GUI, dynamic metadata.
Data Engineering Pipeline
Star Schema lakehouse unifying NoSQL + CSV. Pandas to flatten nested JSON; Parquet + DuckDB for SQL aggregation.
The stack.
The path.
Geospatial Data Scientist
Exp · PlacementEngineering a spatial network of 1.7M+ POIs (Python, NetworkX). Architecting serverless AWS pipelines (S3, Glue, Athena) for high-dimensional geospatial datasets at scale.
MSc Business Analytics
Edu · MScModules in Data Engineering (MLOps, Docker, AWS, SQL), Operational Analytics, Machine Learning, Predictive Analytics.
Data Scientist
Exp · InternshipBuilt a data-scraping pipeline to identify potential client leads, securing 7 startup contracts for employee insurance coverage. Cleaned and processed 120,000+ records.
BSc Business Management — Data Analytics
Edu · BScModules in Data Analytics, Machine Learning, Information Systems. Co-founded the Data Analytics Society (200+ members, 2023–2025).
Foundations deepened.
I aim to learn something new every day. The tech industry moves fast — especially AI — and I make a habit of keeping current: papers, model releases, ship logs, and the open-source conversations between them. These certifications are checkpoints, not endpoints.
Mathematics for Machine Learning
Linear algebra · Multivariate calculus · PCA
Advanced Data Analytics
Statistical methods · Regression · ML modeling
Power BI Desktop
DAX · Power Query · Data modeling
Data Analyst
Python · SQL · Tableau · IBM Cognos
Dimitris Athinaios
dimo.atheneos@gmail.com↗Heavy on AI-assisted development — Claude, Cursor, agents, MCP. For specs I use claude-strike; for diagnosing my AI agents I use Diagnostic by IfixAi. The receipts live at @sirdath.