Іван

Ivan Biloshytskyi
Junior Python Developer · Kyiv, Ukraine · Remote
[відкрити контакти](див. вище в блоці «контактна інформація») · [відкрити контакти](див. вище в блоці «контактна інформація») · github.com/01IVAN10 · [відкрити контакти](див. вище в блоці «контактна інформація»)
SUMMARY
Third-year Software Engineering student at Kyiv Polytechnic Institute. I build async Telegram bots in Python with
payments, webhooks and LLMs, cover them with tests and deploy them. Looking for a remote role, part-time or
full-time with flexible hours.
PROJECTS
trainer-access-bot — paid access bot for a private Telegram channel (demo) GitHub
Python 3.12, aiogram 3, aiohttp, monobank acquiring API, SQLite (aiosqlite), cryptography, pytest, Railway
Full subscription lifecycle: plan → monobank invoice → automatic channel join approval → renewal reminders
→ grace period → removal from the channel.
Bank webhooks verified by ECDSA signature on the raw bytes, with amount checks; a background
reconciliation with the bank API covers webhooks that never arrive.
A payment is credited exactly once even when the webhook, the user's button and reconciliation fire at the
same time (atomic conditional write); the same approach closes the "paid at the moment of removal" race.
217 offline tests with fake Telegram and bank (duplicate webhooks, forged signatures, restarts, races); test
suite strength checked with mutation testing. Deployed on Railway.
barbershop-ai-support-bot — AI support bot with human handoff (demo) GitHub · video
Python 3.12, aiogram 3, Google Gemini (structured output), SQLite, Railway
The model returns {answer, needs_human, reason} and the code decides on handoff; a filter blocks actions
the model invents ("booked you", "I'll pass it on").
Handoff via Telegram group topics: a case card per client, two-way message relay, auto-closing of idle cases.
Eval set of 34 scenarios (incl. prompt injection), each run 3× on the real model: from 88% to 100% (102/102)
after prompt and data fixes. Cost ~$0.00075 per answer.
Crypto Backtesting Framework — backtesting trading strategies on Binance data GitHub
Event-driven backtest engine without lookahead (next-bar execution, fees, slippage, ATR stops), risk sizing as
a pure function; 46 pytest tests on lookahead and stop logic.
Euro 2024 Scout — analytics web app on Streamlit live app · GitHub
Modular pandas/NumPy package; page smoke tests and dataset checks in GitHub Actions; deployed on
Streamlit Cloud.
Teaching materials website — for a school teacher (JavaScript, Cloudflare Pages) uroky-site.pages.dev
EXPERIENCE
Business Analyst (part-time) — AgriAnalytica, agri-fintech Jan 2026 – Jun 2026
Wrote requirements for a marketplace for farmers and a credit-scoring module (user stories, use cases, UML)
and handed them over to developers.
Modelled business processes (BPMN) and built dashboards for management in Python (pandas, matplotlib)
and Excel.
EDUCATION
Igor Sikorsky Kyiv Polytechnic Institute — Faculty of Software Systems and Applied 2024 – 2028 (expected)
Mathematics, BSc in Software Engineering
SKILLS
Python: asyncio, aiogram 3, aiohttp, pandas, NumPy, Streamlit
Testing: pytest, fakes and test doubles, mutation testing, LLM evals
Backend & data: REST APIs, webhooks, ECDSA signatures, Gemini API, SQLite, PostgreSQL, SQL
Other: Git, GitHub Actions, Railway, Streamlit Cloud, JavaScript
Languages: Ukrainian (native), English (B2)