feat(app): NVIDIA Senior Production Engineer, DGX Cloud submitted; on-duty, GPU and networking KB gaps closed; triage 2026-10-22

Application (submitted 2026-10-22, Workday):
- NVIDIA JR2027105, Switzerland Remote. Evidence Fit 65, Stretch (SLIs and
  error budgets not evidenced), hard gate PASS, channel Weak, no cover letter.
- Package in output/NVIDIA_ProdEng_DGX_Cloud/: session, resume .tex, critique
  (Document Quality 85 before Edit 1; Tier 1 + all Tier 2 applied), verbatim JD.
- Prior NVIDIA req JR2024256 was pulled (last listed 2026-09-18; page gone
  2026-10-22): noted on that application, so there was no stacking conflict.

KB (claims.json, user-confirmed 2026-10-22):
- SW-11: current rotating engineer-on-duty role, Opsgenie alerting, alerts and
  bugs auto-create Jira tickets; he writes alert thresholds (e.g. data quality).
- BS-1: the Bosch inference ran on NVIDIA GPUs; CUDA registered forbidden as a
  skill (system context only, little configuration done).
- BS-4: the PoC included Grafana, Prometheus and Loki; ELK Stack, Grafana,
  Prometheus and Loki registered as PoC-context skills.
- networking: certification context only (AWS SA Associate); work uses
  automated landing zones.
- Extraction and CLAUDE.md log row added.

Scout state:
- decisions: RUAG Senior DevOps (Kubernetes) C5I parked (home-office terms);
  SLB Data Scientist skipped (user); NVIDIA JR2008695 skipped (C++/Rust 8y);
  JR2027105 applied; Norway/Apple/Swiss JD-read triage decisions.
- employer_reputation: SLB (Glassdoor 3.9), MazeMap and SpareBank 1
  Nord-Norge (unchecked).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

resume_builder/canonical/claims.json

@@ -363,7 +363,7 @@
         "source": "Bosch Arbeitszeugnis, Dresden 31.12.2022 (primary document, read 2026-09-11); operating constraint user-clarified 2026-09-11",
         "ownership": "individual - integration and operation; model development is NOT his"
       },
-      "scope": "Designed and executed integration of containerized ML inference into a 24/7 semiconductor production environment; model-development ownership is not established. Toolchain includes Docker, Kubernetes and Ansible - Ansible use at Bosch was intensive (user-confirmed 2026-08-27) and extended to configuration management and infrastructure automation beyond this single integration, making it standalone Infrastructure-as-Code evidence. OPERATING CONSTRAINT, user-clarified 2026-09-11: the fab runs continuously, and a defect in the software could in the worst case halt production - an outcome that must never occur. That is a FAILURE-CONSEQUENCE constraint (blast radius of a bug), NOT a deployment-window constraint. Write it as 'a defect could have stopped production' or 'software whose failure would have halted a continuously operating fab'. ZEUGNIS WORDING (primary, 2026-09-11): 'Erarbeitung und Durchfuehrung von Integrationsstrategien fuer den Einsatz von Machine Learning Modellen in der 24/7 Fertigungsumgebung' - devising AND executing integration strategies for ML models in the 24/7 manufacturing environment. This corroborates the integration-and-operation framing and the 24/7 context directly. DESIGN, user-confirmed 2026-10-21 (knowledge_base/extractions/2026-10-21_software_design_architecture_clarification.md): software design and architecture of the ingestion and inference scripts were part of his Bosch work (2020-2022) and count toward a 'years of software design and architecture' requirement. Scope 'designed' to his own scripts and the inference integration - never the fab IT architecture, an architect title or design authority.",
+      "scope": "Designed and executed integration of containerized ML inference into a 24/7 semiconductor production environment; model-development ownership is not established. Toolchain includes Docker, Kubernetes and Ansible - Ansible use at Bosch was intensive (user-confirmed 2026-08-27) and extended to configuration management and infrastructure automation beyond this single integration, making it standalone Infrastructure-as-Code evidence. OPERATING CONSTRAINT, user-clarified 2026-09-11: the fab runs continuously, and a defect in the software could in the worst case halt production - an outcome that must never occur. That is a FAILURE-CONSEQUENCE constraint (blast radius of a bug), NOT a deployment-window constraint. Write it as 'a defect could have stopped production' or 'software whose failure would have halted a continuously operating fab'. ZEUGNIS WORDING (primary, 2026-09-11): 'Erarbeitung und Durchfuehrung von Integrationsstrategien fuer den Einsatz von Machine Learning Modellen in der 24/7 Fertigungsumgebung' - devising AND executing integration strategies for ML models in the 24/7 manufacturing environment. This corroborates the integration-and-operation framing and the 24/7 context directly. DESIGN, user-confirmed 2026-10-21 (knowledge_base/extractions/2026-10-21_software_design_architecture_clarification.md): software design and architecture of the ingestion and inference scripts were part of his Bosch work (2020-2022) and count toward a 'years of software design and architecture' requirement. Scope 'designed' to his own scripts and the inference integration - never the fab IT architecture, an architect title or design authority. HARDWARE, user-confirmed 2026-10-22 (knowledge_base/extractions/2026-10-22_oncall_networking_gpu_clarification.md): the inference ran on NVIDIA GPUs, with CUDA used for the performance boost - his words: 'The ML inference did ran on NVIDIA GPUs (using CUDA for performance boost - but i never did much configuration there)'. Writable: 'GPU-backed inference containers on NVIDIA GPUs' as a description of what he integrated and deployed. CUDA is a property of the system, not his skill: he did little CUDA or GPU configuration, so never claim CUDA work, GPU configuration, GPU tuning or GPU performance analysis.",
       "allowed_verbs": [
         "integrated",
         "containerized",
@@ -375,7 +375,12 @@
         "trained the image classification model",
         "owned the full ML lifecycle",
         "a production line that could not be stopped for a deployment",
-        "could not be stopped for a deployment"
+        "could not be stopped for a deployment",
+        "CUDA programming",
+        "configured CUDA",
+        "GPU performance tuning",
+        "optimized GPU performance",
+        "GPU performance analysis"
       ],
       "metrics": "qualitative reduction in manual classification is source-backed; exact amount unverified"
     },
@@ -414,7 +419,7 @@
         "source": "Bosch Arbeitszeugnis, Dresden 31.12.2022 (primary document, read 2026-09-11)",
         "ownership": "individual - proof of concept"
       },
-      "scope": "Built an ELK/Kafka anomaly-detection proof of concept and monitoring; not a company-wide observability platform. ZEUGNIS WORDING (primary): 'Durchfuehrung eines POCs fuer das Logfile-Management mit ELK-Stack' - the PoC framing is confirmed by the document itself, not just by policy.",
+      "scope": "Built an ELK/Kafka anomaly-detection proof of concept and monitoring; not a company-wide observability platform. ZEUGNIS WORDING (primary): 'Durchfuehrung eines POCs fuer das Logfile-Management mit ELK-Stack' - the PoC framing is confirmed by the document itself, not just by policy. STACK, user-confirmed 2026-10-22 (knowledge_base/extractions/2026-10-22_oncall_networking_gpu_clarification.md): besides ELK and Kafka, the PoC included monitoring and alerting with GRAFANA, PROMETHEUS and LOKI. That was previously recorded only in the user's earlier CV (thiessen_cv_master_profile.md), not in the Zeugnis, which names the ELK log-management PoC. All of it stays PoC-scoped: never a production monitoring platform.",
       "allowed_verbs": [
         "built",
         "implemented",
@@ -819,6 +824,36 @@
         "formal mentoring programme - not established"
       ],
       "metrics": "none; number of junior developers and outcomes are not recorded"
+    },
+    {
+      "id": "SW-11",
+      "evidence": {
+        "source": "user confirmation 2026-10-22; knowledge_base/extractions/2026-10-22_oncall_networking_gpu_clarification.md",
+        "ownership": "team-role - member of a rotating engineer-on-duty role; individual - alert thresholds for his own pipelines"
+      },
+      "scope": "CURRENT on-duty and alert handling at Swisscom (user answer 2026-10-22 to 'do you monitor or get alerted on your Glue/Airflow pipelines; is there on-call in the current unit?'). Verbatim: 'Yes, we get alerted ... we are using opsgenie and we have a engineer on duty role which rotates. Bugs and Alerts also automatically create Jira issue tickets which we dann asign an fix.' Established: alerting through OPSGENIE; a ROTATING ENGINEER-ON-DUTY role in his team, which he takes part in; bugs and alerts automatically create Jira tickets that the team assigns and fixes. PHASE: current (asked about the current unit, ART DPT); never date it. This is the current counterpart of the past SW-2 on-call (Rufbereitschaft) at ART B2B. NOT established: who built or configured the alerting, the Opsgenie routing or the Jira automation (do not claim he did); which metrics or logs feed the alerts; SLIs, SLOs or error budgets; any incident count, response time, MTTR or availability figure; whether the duty covers nights and weekends (do not write 24/7). ALERT DEFINITIONS, user-confirmed 2026-10-22 (follow-up 'do you also write or adjust the alerts for your own pipelines?'): 'yes, i wrote e.g. thresholds for data quality etc....' So he defines alert thresholds for his own pipelines, data-quality thresholds being the named example. Writable: 'defined alert thresholds, including data-quality checks, for my pipelines'. Still NOT established: other alert types (latency, freshness, cost) - do not list them; the Opsgenie routing and the Jira automation, which he did not set up.",
+      "allowed_verbs": [
+        "serve",
+        "take part in",
+        "respond to",
+        "triage",
+        "fix",
+        "resolve",
+        "define",
+        "set"
+      ],
+      "forbidden": [
+        "configured Opsgenie",
+        "set up Opsgenie",
+        "designed the on-call process",
+        "owned incident management",
+        "SRE role or title",
+        "24/7 on-call",
+        "any incident count, MTTR, response-time or availability figure",
+        "built the alerting platform",
+        "set up the Opsgenie or Jira integration"
+      ],
+      "metrics": "none"
     }
   ],
   "skills": [
@@ -1273,6 +1308,53 @@
       "evidence": "production-historical",
       "output": "allowed",
       "note": "Registered 2026-10-14 from existing canonical claim scopes - no new user statement, same procedure as Jenkins and GitLab CI/CD on 2026-10-08. Was evidenced in scopes but absent from this skills list, so a skills line could not cite it (surfaced by the BIS jr100498 package). BS-2 (Bosch 2020-2022): data services over Hadoop/Impala for fab analysis teams. Query and data-service use; does not establish Hadoop cluster administration or Cloudera administration."
+    },
+    {
+      "name": "Opsgenie",
+      "evidence": "hands-on-current",
+      "output": "allowed-with-context",
+      "context": "SW-11: alerting for his team's rotating engineer-on-duty role at Swisscom.",
+      "note": "User-confirmed 2026-10-22. Use as a responder, never as its administrator or the author of the alert routing."
+    },
+    {
+      "name": "CUDA",
+      "evidence": "system-context-only",
+      "output": "forbidden",
+      "note": "User-confirmed 2026-10-22: the Bosch inference (BS-1) ran on NVIDIA GPUs with CUDA for the performance boost, but he 'never did much configuration there'. CUDA describes that system, not a skill he has. Registered forbidden so the scanner catches it on documents; the honest wording is 'GPU-backed inference on NVIDIA GPUs' (BS-1). In an interview he can say the models used CUDA and that his part was integration and deployment."
+    },
+    {
+      "name": "networking",
+      "evidence": "certification-and-personal",
+      "output": "certification-context-only",
+      "note": "User answer 2026-10-22 to 'any AWS VPC, security-group or private-endpoint work at Swisscom, or firewall, DNS, proxy or load-balancer work at Bosch?': 'No, not really - probably only my AWS solution architect cert covers AWS network topics... but day to day work is using automated landing zones.' So: no professional network configuration. AWS networking is covered by the AWS Certified Solutions Architect - Associate (26 Sep 2024, active; thiessen_certifications.md #7), certification context only. At work he USES automated AWS landing zones provided by the platform - never claim he built or designed landing zones, VPCs or network topology. Personal: PP-3 runs nginx, a VPN server and a mail server (personal label required; PP-3 forbids datacenter or physical-infrastructure claims). JD lines on networking fundamentals stay Adjacent at best."
+    },
+    {
+      "name": "ELK Stack",
+      "evidence": "proof-of-concept",
+      "output": "allowed-with-context",
+      "context": "BS-4: Bosch log-management and anomaly-detection proof of concept (2020-2022).",
+      "note": "Registered 2026-10-22. ELK named in the Bosch Zeugnis. Always keep the PoC label; never imply production or platform ownership."
+    },
+    {
+      "name": "Grafana",
+      "evidence": "proof-of-concept",
+      "output": "allowed-with-context",
+      "context": "BS-4: Bosch log-management and anomaly-detection proof of concept (2020-2022).",
+      "note": "Registered 2026-10-22. User-confirmed 2026-10-22; previously only in the user's earlier CV. Always keep the PoC label; never imply production or platform ownership."
+    },
+    {
+      "name": "Prometheus",
+      "evidence": "proof-of-concept",
+      "output": "allowed-with-context",
+      "context": "BS-4: Bosch log-management and anomaly-detection proof of concept (2020-2022).",
+      "note": "Registered 2026-10-22. User-confirmed 2026-10-22; previously only in the user's earlier CV. Always keep the PoC label; never imply production or platform ownership."
+    },
+    {
+      "name": "Loki",
+      "evidence": "proof-of-concept",
+      "output": "allowed-with-context",
+      "context": "BS-4: Bosch log-management and anomaly-detection proof of concept (2020-2022).",
+      "note": "Registered 2026-10-22. User-confirmed 2026-10-22; previously only in the user's earlier CV. Always keep the PoC label; never imply production or platform ownership."
     }
   ],
   "global_forbidden_output_patterns": [

resume_builder/reference/current_applications.md

@@ -5,7 +5,7 @@ Use current outcome and next action below. Imported legacy notes are historical,
 
 | Application | Current outcome | Interview reached | Next action |
 |---|---|---|---|
-| NVIDIA - Senior DevOps Engineer, Zürich (req JR2024256) | applied | not recorded | Await response. At first contact, clarify homeoffice/onsite requirements and total package. Baseline is 3 homeoffice days/week; fewer days require a clear overall improvement. Note: `claims.json` has no shell-scripting entry, so **Bash is omitted** from Skills although the JD names it and ten years of Debian admin plus Ansible work plainly involve it - same shape as the Ansible/Linux gaps found 2026-08-27. Confirm with the user and record it. Groovy stays out regardless. |
+| NVIDIA - Senior DevOps Engineer, Zürich (req JR2024256) | applied | not recorded | No action expected; req pulled ~2026-09-20. Treat as dormant. Does not block NVIDIA JR2027105 (Senior Production Engineer, DGX Cloud). |
 | RUAG C5I - AI Engineer C5I, Thun (application ID 18027) | rejected_after_interview | yes | Abgeschlossen: Absage nach Interview; Anforderungen vor allem beim Homeoffice nicht vereinbar. Keine weitere Aktion fuer diese Bewerbung offen. |
 | Schweizer Armee - Kommando Cyber (Kdo Cy), DevOps Engineer III (Data Platform), Zimmerwald (Ref JRQ$540-19848) | rejected_no_interview | not recorded | Closed; no pending application action. Read recorded history before reopening. |
 | Citadel Securities — Platform Engineer, Research Platform Engineering, Zürich (also NY/Miami) | applied | not recorded | Await response. At first contact, confirm whether the application is for the Zurich seat; relocation to the US remains ruled out. Clarify homeoffice/onsite requirements and the Zurich compensation package. Baseline is 3 homeoffice days/week; fewer days require a clear overall improvement. The posted US salary range is not a Zurich figure. |
@@ -36,6 +36,7 @@ Use current outcome and next action below. Imported legacy notes are historical,
 | Kongsberg Discovery (Seatex) — Lead Engineer, DevOps and Test Infrastructure, Trondheim | applied | not recorded | Wait for Kongsberg response. Screen prep: embedded targets/HIL question, recency of DevOps/test work (2015-22), relocation to Trondheim, clearance eligibility as German citizen resident in Switzerland, on-site expectations. Package: output/Kongsberg_Lead_Engineer_DevOps_Test_Infra/. |
 | BIS (Bank for International Settlements) — Senior/Lead Data & Analytics Engineer, MED IT, Basel (jr100498, 3-yr term) | applied | not recorded | Await BIS response. If contacted: ask office days and total package; weigh the 3-year fixed term against the permanent Swisscom post. Interview prep is in the critique: SDMX answer, Swisscom dashboard tool (not in KB), attributed schedule example for project planning, Lead story. |
 | Google — Staff Software Engineer, Shopping Graph (Feeds Infra), Zürich | applied | not recorded | Await Google response. Google limit: next Google application window counts this submit (3 per rolling 30 days). At first contact clarify level (L6 vs L5), Zurich hybrid days vs current 3 homeoffice days, and total package. Interview prep: SW-4 prototype story as the technical-direction example, plus one concrete Staff-level example the user names; Java currency; end-to-end system design with scale and trade-offs. Package: output/Google_Staff_SWE_Shopping_Graph/. |
+| NVIDIA — Senior Production Engineer, DGX Cloud, Switzerland Remote (JR2027105) | applied | not recorded | Await response. Send the LinkedIn note to an NVIDIA recruiter or DGX Cloud engineer (draft in session). Interview prep: incident debugged end to end (SW-2/SW-11); safe model rollout in the fab (BS-1, rollout mechanics not in KB); Kubernetes recency; distributed systems at scale; vLLM/SGLang/Dynamo not used. Package: output/NVIDIA_ProdEng_DGX_Cloud/. |
 
 ## Current targeting notes