Gate 01
History / CV
- About
- Experience
- Education
Hi. I am Iron, your portfolio guide.
AI / Data / Product Engineering
I build practical systems at the intersection of product engineering, data workflows, automation, and applied AI. My current focus is MLOps, AI Engineering, LLMOps, data infrastructure, and decision-support products.
Gate 02
Data ML AI
- MLOps
- AI Engineering
- Data Workflows
- Kaggle
Gate 03
Books Blog Fun
- Writing
- Reading
- Experiments
Project Dossiers
Research, legal AI, energy analytics, packages, and interface systems - each with a compact proof trail.
Gate 04product
Legal AI workflow
AMCU Procurement Copilot
A production-oriented demo MVP for AMCU/Prozorro matters: it imports tender evidence, resolves public legal sources, runs structured AI steps, and keeps every draft inside a lawyer-review boundary.
- Partner-review package: intake, issues, evidence gaps, argument matrix, draft complaint, memo, red-team critique, citations, and trace.
- Public-source routing across Prozorro, AMCU/Rada, complaint candidates, and official EDRSR open-data court metadata.
- Strict schema validation and unsupported-claim marking prevent the AI output from pretending weak evidence is established fact.
Gate 05research
MSc Data Science thesis
Smart Energy Arbitrage
A defended energy-arbitrage system with Dagster/FastAPI/Postgres, a Nuxt operator dashboard, deterministic LP/V2+ evidence, and bounded DT/HF shadow comparisons. The boundary is explicit: preview/read model only, no bids and no autonomous dispatch.
- Public Ukraine BESS Arbitrage Index turns thesis work into a source-backed portfolio/data product.
- Operator preview separates official DAM/IDM rows, forecast-store rows, readiness, strategy evidence, and no-execution controls.
- Research evidence includes LP/oracle, V2+, DT safe-switch, and HF value-aligned shadow lanes with documented claim boundaries.
Gate 06collection
Selected course finals / 2025-26
AI / Data Coursework Lab
A focused archive of recent coursework with enough depth to stand behind: controlled ML evaluation, statistically grounded product experimentation, and production-shaped streaming and batch data pipelines.
- Three substantial final-course projects selected from the last two years, not a dump of every homework repository.
- Covers controlled AI/ML evaluation, product decision-making, and practical data infrastructure.
- Each tab carries its own result trail, technical stack, and source repository.
Gate 07package
Published Vue package
@fulliron/fluid-text-vue
A packaged version of the portfolio title system: WebGL smoke/fluid text, composable lifecycle pieces, typed options, and a build path that can be reused outside this site.
- Extracted from the arcade portfolio into a reusable npm package.
- Keeps shader/runtime controls typed and componentized for safer reuse.
- Built alongside the site so the portfolio demonstrates the package in context.
Gate 08repository
This site
Full Iron Arcade Portfolio
A deliberately memorable portfolio shell: pixel-art navigation, multilingual content, responsive station modals, animated guide characters, and a content model that can grow with real projects instead of static resume blocks.
- Private source repo with branch-based review, visual QA screenshots, typecheck, build, and audit gates.
- Custom pixel sprites and fluid title rendering support a distinct but still functional product surface.
- The second screen now behaves like a project archive instead of repeating the first screen islands.
Gate 09repository
Browser graph exploration
Brancher.io extension work
Interactive browsing-history graph work that sits between product UX, data visualization, and browser-extension thinking - useful as an early proof of interface-heavy graph exploration.
- Demonstrates the same throughline as current work: make hidden workflows visible and explorable.
- Useful positioning bridge between older frontend/product work and current data/AI interfaces.
- Kept as a directional artifact while newer public repos carry the deeper technical proof.