I help teams build resilient automation and data systems that remain testable, observable, and recoverable under real-world failures.
Senior .NET/Python engineer with 15+ years of professional C# experience. My public work spans browser automation, developer tooling, applied ML, and reproducible data pipelines.
Available for part-time consulting and project-based work in:
- resilient browser and API automation
- developer tooling for failure and recovery testing
- reproducible data and applied-ML pipelines
I am not currently considering full-time roles.
- Email: bockuden@gmail.com
- Telegram: @bockuden
A ready-to-run deterministic failure sandbox for Playwright, Selenium, scrapers, and HTTP automation workers.
It provides reproducible browser and API scenarios for transient 503
responses, permanent failures, delays, authentication, pagination, duplicate
records, DOM changes, cancellation, and checkpoint recovery.
Each run is isolated by run_id, allowing the same failure sequence to be
reproduced locally and in CI.
The project is distributed as a Python package and container image. It includes a CLI, a stable OpenAPI contract, automated compatibility checks, and runnable resilience examples.
Repository · PyPI · Container · Resilience Challenge
A production-style .NET 10 and Playwright worker designed for interrupted jobs, repeated delivery, transient failures, DOM changes, duplicates, cancellation, and recovery from durable checkpoints.
The project includes:
- SQLite persistence and idempotent processing
- bounded retries, concurrency, and per-target rate limiting
- checkpoint-based resume after interruption
- structured logging and OpenTelemetry metrics
- screenshots, HTML snapshots, traces, and machine-readable failure evidence
- unit, integration, and deterministic browser E2E tests
- a reproducible Docker Compose demo
The worker is validated against a pinned release of the independently published Resilient Automation Test Stand.
Repository · Release · Architecture · Compatibility matrix
I built a reproducible, resource-aware PCAP analytics pipeline with leakage-aware validation, transparent model comparison, resumable data processing, and safe local scoring.
FlowSentinel converts multi-gigabyte CICIoT2023 captures into auditable packet-feature datasets, keeps TCP flows disjoint across train, validation, and test, and supports both supervised intrusion classification and benign-only anomaly scoring. The published benchmark compares tabular, neural, and unsupervised baselines without claiming unseen-capture generalization where the source data cannot support it.
Repository · Release · Benchmark · Data card
My research repositories emphasize explicit assumptions, evidence provenance, reproducible pipelines, and honest reporting of negative or inconclusive results.
| Project | Research question | Approach | Current conclusion |
|---|---|---|---|
| OpenAI 30 GW U.S. Grid Exposure Screen | How sensitive could U.S. balancing authorities be to different hypothetical allocations of 30 GW of compute load? | EIA data, scenario analysis, evidence ledger, data-quality audits, robustness checks, and reproducible offline builds | A dispersed load is modest relative to combined national peak, while concentrated deployment can create materially larger local exposure. This is an independent sensitivity study, not a siting or blackout forecast. |
| US Fertility, Housing, and Income Drivers | How are housing costs, housing structure, income, and fertility associated across different levels of U.S. data? | State-year baseline, metro fixed-effects models, and SIPP household-panel analysis | Housing-related signals remain visible in richer panel designs, but the current evidence does not justify a simple causal claim. |
| Crypto Cointegration Stability Study | Do highly correlated crypto perpetual pairs remain cointegrated and economically useful out of sample? | Rolling correlation, Engle–Granger tests, stability filters, walk-forward evaluation, and trading-cost modeling | Most candidates are unstable or unprofitable. The limited positive result is exploratory and requires confirmation on untouched data. |
- 15+ years of professional C# development
- Browser automation and reliable web data extraction
- Python, FastAPI, Playwright, and SQLite
- Resilient workers, retries, idempotency, and checkpoint recovery
- OpenCV, TensorFlow, and applied machine learning
- Time-series analysis, econometrics, and quantitative research
- Reproducible data pipelines and research artifacts
- University degree in education, specializing in physics and mathematics teaching