1
Curious builder0 XP earned · 300 to level 2
0 daysFinish a lesson to begin
Badge collection0 of 6 unlocked
51 small wins to finish your pathNext question →
Your demo worked perfectly, but in the customer's environment nothing works. Where do you look?
30-second answerSay your answer out loud first, then reveal.
Debugging checklist
| Layer | Typical issue | Quick test |
|---|---|---|
| Network | Corporate proxy, TLS inspection breaking certificates, LLM endpoint blocked | curl the endpoint from the server; check proxy env vars and CA bundles |
| Identity | Service account lacks permissions; SSO tokens expire | Call the data API with the same identity; inspect 401/403s |
| Data | Real files 10x larger, different templates, non-UTF-8, Hindi text | Run the pipeline on 10 real samples; log parse output |
| Config | Wrong region endpoint, missing secrets, different model name/version | Print effective config at startup (redacted) |
| Dependencies | Different Python/library versions, missing OCR binaries | Pin versions; containerise |
| Limits | Gateway timeouts at 30s, memory limits, provider rate limits | Check the logs for timeouts and 429s |
| Time / locale | Date formats, time zones | Unit tests with their formats |
Prevention for next time
- Containerise and use infrastructure-as-code.
- Do a "walking skeleton" deployment into their environment in week 1 (even a hello-world that calls the model and reads one document).
- Use a pre-engagement environment checklist (network allow-lists, accounts, quotas).
Related
- Previous: Q20. Midway through the project, the customer's security team blocks sending any data to external LLM APIs. What are your options?
- Next: Q22. Business wants to launch fast, IT wants a full architecture review, and compliance wants more controls. Stakeholders disagree. How do you move forward?
Little by little, you're building something great.