Back to case studies
2022-2024Commercial data tooling
Resilient data collection workflows
Collection and debugging workflows for external web systems where behavior changes and failures must be diagnosable.
Diagnostics
HTTP + browser runtime
Reusable logic
Debuggable collection
Changing targets need diagnostics, not just parsers.
The goal is to make failures explainable enough to fix.
1Request and browser behavior inspected together
2Failure cases turned into reusable checks
3Parsers designed around real target behavior
Python Data / Backend Developer
PythonWeb scrapingReverse engineeringPlaywrightClickHouse
Problem
External targets changed often, and failures were hard to reproduce from a simple error message.
Solution
Worked with request tracing, browser automation, parsers, diagnostics, and reusable collection logic.
Impact
Failures became easier to classify, reproduce, and fix without starting from zero each time.
What I did
- Analyzed HTTP and JavaScript behavior for changing external systems.
- Built and adjusted collection logic around real target behavior.
- Improved diagnostics so failures were easier to reproduce.
What it shows
- A parser is only useful if the failure path is visible.
- Data collection work rewards patience with edge cases.
Related work
More projects
2025-2026
nnzen model catalog
A live LLM catalog that collects model data, normalizes it, and makes model comparison faster.
PythonFastAPILLM APIsRAGVector DB / pgvector
2025
MCP core for an LLM assistant
Backend core for an LLM assistant with plugin execution, hot reload, tool chains, and explicit context handoff.
PythonFastAPIMCPTool callingLLM APIs