Vadym Arnaut

Vadym Arnaut

Software engineer · Indianapolis, Indiana

I build software where being exact matters.

Full-stack engineer working across React, TypeScript, Python and Postgres. I ship products end to end and contribute fixes to open source.

Open to full-stack and forward-deployed roles

Equip

In production2025 – now

A free learning platform for Bible schools and volunteer-run programs. Teachers build a course once; students take it in their own language.

Equip landing page: “Study Scripture in order, not in fragments”
Fig. 1 equipbible.com, October 2026
  • Bilingual content is symmetric by design: every translatable field is declared in one registry, and CI fails if any layer drifts from it.
  • Scripture quotes are matched against canonical texts and substituted, so the translation model never paraphrases a verse.
  • Multi-tenant on Supabase with row-level security; 76 policies rewritten in one migration to remove an init‑plan performance trap.
Stack
React, TypeScript, FastAPI, SQLAlchemy 2, Supabase, Vercel

History Globe

In production2026

A globe of the ancient world with a time slider from 3500 BC to AD 1300. Borders shift year by year, cities rise and fall, and biblical places link to the verses that name them.

History Globe showing the Levant in 1003 BC
Fig. 2 The Levant in 1003 BC
  • 1,200+ places on the map; 1,100+ carry the verse and verbatim quote that name them, and the build fails if the evidence does not match.
  • 45+ guided journeys, from Abraham to the seven churches, each checked verse by verse against the KJV and Synodal texts.
  • Data comes only from sources that allow reuse, with automated guards that reject restricted coordinates.
Stack
MapLibre GL, React, TypeScript, Python, Postgres, Vitest

TanStack AI

Open source2026

A type-safe, provider-agnostic AI SDK for TypeScript. I fix bugs in its streaming core and in the Anthropic and Ollama adapters.

  • Restored provider tools and tools-plus-structured-output for the newest Claude models; shipped in @tanstack/ai-anthropic 0.18.13. #1423
  • Traced why multi-turn structured output broke only on Anthropic, and locked the fix down with two mutation-tested guards. #1472
  • Made the stream processor run client tools that a finished run from another AG-UI server leaves pending. #1543
  • Reported token usage when an Anthropic stream stops at max_tokens, so callers can see what a cut-off answer cost. #1609
Stack
TypeScript, Anthropic API, Vitest, Playwright
  • HTTP QUERY Lab

    Sent the new HTTP QUERY method (RFC 10008) through real clients, frameworks and CDNs to see what survives.

  • YouVersion usfm-references

    Added a book-name-to-USFM converter and numbered-book prefix normalization.

  • FinScope

    Personal finance app with machine-learning insights. My bachelor thesis.

  1. 2025 – 2026

    Software EngineerInvisible Technologies

    Core engineer on a B2B app where clients review computer‑vision output and generate reports. Cut query time by 87% on ~100M-row tables through indexing, query rewrites and connection-pool tuning.

  2. 2023 – 2025

    Project ManagerNew Caliber Construction

    Ran renovation jobs from estimate to handover, and automated the billing and reporting behind them.

  3. 2021 – 2025

    B.Sc. Computer ScienceBorys Grinchenko Kyiv Metropolitan University

    Thesis: FinScope, a full-stack personal finance app with machine learning, defended with the highest grade.

Email is the fastest way to reach me.

arvavitofficial@gmail.com