Олександр

OLEKSANDR DOVHULIA
Systems & Data Analyst — data integrity, validation, experiment design
Kyiv, Ukraine (Remote) • [open contact info](look above in the "contact info" section) • github.com/OleksanderSS

SUMMARY
Systems analyst with an M.Eng. in quality and safety engineering, five years as a pre-match and live sports trader at Favbet pricing markets against a
margin target where one wrong quote is exploited within minutes, and over a year directing and building market-data research pipelines. What I do best
is find where data lies — look-ahead leakage, survivorship bias, silently dropped records, prices on the wrong basis — and prove each fix with a test
that fails on the old code and passes on the new one.
EXPERIENCE
Independent Research — Market-Data Pipeline & Validation Framework Sep 2025 – Present
Python, DuckDB, pytest, Git, FastAPI; AI-assisted development (Claude Code, Gemini, Codex) — architecture, test design, implementation, and code review
• Remote
• Directed and built a 7-stage pipeline (prices, SEC filings, FRED macro, news, options, analyst estimates) across 1,226+ commits, owning system
architecture, implementation logic, and acceptance testing.
• Look-ahead audit: found one defect shape in three separate data sources: a date stored without a time of day, read as midnight. For SEC 10-K/10-Q
data, 46% of facts were visible to the model before EDGAR had accepted the filing; fixed by keying on the acceptance timestamp (99.98% matched,
control 99.87%).
• Survivorship bias: showed that the hindsight-chosen universe of 105 stocks, with no model at all, beat the market by 6.9% a year (Sharpe 1.22, t =
6.3, 1996–2023) — the baseline bar any predictive signal had to clear first.
• Silent data loss: identified a cleaning step that deleted every bar beyond 3σ — 14,598 real price moves, including crisis days and earnings releases
— as well as 308 classification models trained and scored as regressions.
• Cost modeling & correction: per-share commission had been priced on split- and dividend-adjusted prices (AAPL 2013: 32× below the traded price),
overstating it about five-fold (9.7 → 2.0 bp per turnover on rows matched to real traded prices); fixed the collector to retain the traded price and
updated the published benchmark. With realistic friction, documented that no tested strategy outperformed a passive benchmark.
• Verification layer: engineered nightly unattended test suites running 900+ contract tests with a no-new-failures ratchet; every measurement
pre-registered (rule, forecast, and known-answer control committed before execution), tracking model hit rates alongside empirical outcomes.
Sports Trader (Bookmaker) • Favbet 2020 – 2025
• Priced and managed pre-match and in-play (live) betting lines against defined turnover-margin targets; in tennis, aligned lines in real time with global
market-maker benchmarks.
• In esports, acted as the primary market maker: compiled proprietary match statistics and engineered pricing models from scratch for partner-studio
events with no external feed; the resulting data stream was licensed to external B2B bookmakers.
• Promoted on margin targets and zero critical pricing remarks; caught mirrored odds and pricing latency before they could be exploited.
Quality Department • Fozzy Group 2020
• Incoming quality control, storage checks, and documentation audits under corporate standards and HACCP; sampling and failure analysis (FMEA);
non-conformance reports.

PORTFOLIO
• Market-data audit (github.com/OleksanderSS/market-data-audit): Validation framework and tests proving the integrity findings above, including
pre-registered forecasts, baseline controls, and documented edge cases.
• Lead processing MVP (github.com/OleksanderSS/lead_pipeline_mvp): FastAPI webhook that normalises landing-page payloads, generates an
LLM summary (Gemini), scores leads (HOT/WARM/COLD) via deterministic rules, syncs to Google Sheets, and pushes Telegram alerts; 23 tests, CI
on every push.

SKILLS
• Core: Data validation & leakage audits • Point-in-time architecture • Survivorship bias mitigation • Corporate actions (splits, dividends) • Experiment
design & pre-registration • Known-answer controls • FMEA
• Technical: Python • SQL (DuckDB) • pytest • FastAPI • REST APIs • Git & CI workflows • AI-assisted engineering (Claude Code, Gemini, Codex)

EDUCATION & LANGUAGES
M.Eng., Systems Quality & Safety Engineering — National University of Food Technologies, Kyiv 2015 – 2020
Languages: Ukrainian (Native) • English (Professional Working Proficiency / B2)

Oleksandr Dovhulia — Systems & Data Analyst Resume Confidential & Professional Reference