# Critique: Microsoft — Senior Software Engineer, Industry Solutions Engineering (ISE), Zürich (req 200040836)

**Resume File:** `output/Microsoft_ISE_Senior_SWE/e2e_microsoft_ise_resume.tex` (2 pages, compiled clean)
**Cover Letter File:** `output/Microsoft_ISE_Senior_SWE/e2e_microsoft_ise_cover_letter.tex` (1 page, 299 words)
**JD:** `output/Microsoft_ISE_Senior_SWE/JD_microsoft_ise.txt` — verbatim live fetch 2026-07-03 via Eightfold PCSX API (JD integrity: PASS)
**Date:** 2026-07-03 · **Pass:** 2 (lens reused from Pass 1 — do not rebuild)

---

## Changes Since Pass 1 (applied 2026-07-03, user-directed)

Resume only — CL untouched (passed all checks in Pass 1). Compile verified: 2 pages, summary 5 lines (no orphan), all skills lines single-line, page break identical to baseline.

1. **T1-1:** AI & ML in Production line now reads "…MLOps, model evaluation, data-quality & performance monitoring" — JD's RQ triple complete verbatim.
2. **T1-2:** Software & Data Engineering line 3 now reads "Open-source stack: Apache Kafka, Apache Airflow, PySpark / Spark, Hadoop/Impala; batch & streaming" ("ingestion" dropped for line fit; no contribution claim made).
3. **T2-1/2/3 (single summary rewrite):** "…for agentic AI and LLM workloads, run Python services on Kubernetes, and carry on-call duty as Component Owner." — adds LLM to the recruiter window and binds on-call to the Component Owner role (precision fix); "I work embedded in cross-functional customer and stakeholder teams." — adds the JD's pod descriptor. Tail clause "platform co-ownership, workshops, training" removed to hold the 5-line budget; that signal remains verbatim in BS-3 ("Co-owned the TIBCO Spotfire analytics platform… training engineering users") and GN-1.
4. **T2-4 (conditional metric): NOT applied** — KB check of `experience_bosch.md` / `experience_swisscom.md` found no verified quantitative metric for BS-1/SW-1 beyond 24/7 and 300mm (already used). Adding one would violate anti-fabrication.

### Re-scored dimensions (others unchanged)
| Dimension | Pass 1 | Pass 2 | Why |
|---|---|---|---|
| ATS Keywords (15%) | 8.0 | 9.0 | 20/20 — "model evaluation", "open source", "cross-functional" all present; remaining items semantic-only by nature |
| Summary (10%) | 8.5 | 9.0 | LLM in the 10-second window, cross-functional added, on-call claim probe-proofed |
| Skills (10%) | 8.5 | 9.0 | RQ vocabulary complete, open-source identity named, still zero fabrications |

**Score trajectory: 83.3 → 85.8/100.** At/above the 85+ submit band; theoretical max ~86.5 (residual: no hard metrics in KB, page-2 fill exception). Hard ceiling ~88 unchanged (Azure, OSS contributions — not resume-editable).

**Interview likelihood (updated):** ATS 95% · Recruiter 80% · HR 80% (RQ triple now checkbox-verbatim) · HM 65% (open-source affinity + model-eval line remove two objections) · Panel 70% (unchanged — depth questions remain; bridge points in Part 6 cover them).

---

## Part 1: Domain-Specialist Lens

### Reviewer Persona
ISE engineering manager or principal SWE in EMEA running the hiring-manager screen for a Zürich IC4 req. Daily work: scoping customer engagements, unblocking cross-functional pods (SWE + DS + TPM + design), reviewing code on customer repos in whatever stack the customer runs. Has read 50+ CVs for this posting. Eye-rolls at: buzzword AI claims with no operational substance ("passionate about GenAI"), tool soup, solo-hero claims over org-scale systems, consultants who architect but don't code. Genuinely impressed by: production evidence under real constraints, breadth across stacks ("not just Microsoft" is in their own JD), customer-embedded engineering, scope-honest claims, and — for this specific req — German + already-in-Switzerland (no visa, no relocation).

### Company Context
ISE is Microsoft's global non-billing co-engineering org inside MCAPS: engineers write production code side-by-side with strategic customers' engineers, then feed patterns back into Microsoft products and open source. Their public Engineering Fundamentals Playbook codifies the working style (code-with, testing, observability, agile ceremonies). Zürich req + "German beneficial" → Swiss/DACH strategic accounts (banks, pharma, industrials). Strategic backdrop: USD 400M Swiss datacenter expansion (June 2025) with in-country data residency — regulated Swiss enterprises are the customer base. Recent ISE publication themes: enterprise RAG with document-permission propagation (Entra ID → AI Search), coordinator-based multi-agent architectures, vector/hybrid search evaluation.

### JD Vocabulary Extraction (top 10, ranked)
| # | JD Term | Freq/Placement | Meaning at ISE | Resume Match? |
|---|---|---|---|---|
| 1 | deploying and operating AI systems in production | RQ line | Not research demos — AI running for real customers, operated | SEMANTIC (ML inference in 24/7 fab; "Production AI Foundations" headline) |
| 2 | LLM-based solutions, prompt engineering, RAG | RQ line | Hands-on LLM integration in enterprise settings | YES verbatim (LLM, RAG, prompt engineering in skills) |
| 3 | model evaluation, data quality, performance monitoring | RQ line | Can you tell if the AI solution works and keeps working | PARTIAL — data quality + performance monitoring YES; **model evaluation ABSENT** |
| 4 | modern software engineering practices | Overview + RQ | Playbook fundamentals: CI/CD, code review, testing, design docs | SEMANTIC (CI/CD, code review, TDD, quality gates) |
| 5 | cloud technologies / cloud-based solutions | 4× | Azure-centric but explicitly polyglot org | YES (AWS named honestly, cloud-native) |
| 6 | DRI, on call, monitor/restore, playbook | Responsibilities | Production ownership culture | SEMANTIC-STRONG ("on-call SLA", "incident response and restoration") |
| 7 | cross-functional team (SWE + DS + TPM + designers) | Overview | Pod structure of every engagement | **NO verbatim** (semantic: stakeholder/product-owner partnering) |
| 8 | open source | 3× | Contribution + "not just Microsoft" identity | **NO** (stack is open-source but never named as such) |
| 9 | C, C++, C#, Java, JavaScript, Python | RQ + PQ | Polyglot flexibility, meet customers in their language | YES — 5 of 6 named |
| 10 | observability, availability, reliability | Responsibilities | Operations-at-scale mindset | YES verbatim (skills group name) |

### Domain Vocabulary Map
| Resume Currently Says | Should Say for This JD | Why |
|---|---|---|
| data-quality & performance monitoring | model evaluation, data-quality & performance monitoring | Completes the JD's RQ triple verbatim; backed by IBM AI Engineering cert + operating a production classifier |
| Apache Kafka, Apache Airflow, … (unlabeled) | Open-source data stack: Kafka, Airflow, Spark… | JD says "open source" 3×; his stack IS open source — free honest keyword + "not just Microsoft" culture signal |
| customer and stakeholder teams | cross-functional customer and stakeholder teams | JD's pod descriptor; true of his PO/analyst/domain-team work |
| agentic AI workloads (summary) | agentic AI / LLM workloads | Puts "LLM" in the recruiter's 10-second window, not just skills |

### Gap Ranking
- **Fatal:** none. Every RQ line is covered or truthfully bridgeable.
- **Serious:** (1) "model evaluation" — RQ-line term, currently absent (bridgeable: JD asks only *understanding*); (2) Azure — competitors are Azure-native; resume/CL handle it honestly (AWS named, transfer argued) but it remains a real gap; (3) open-source *contributions* — JD says "contribute to open source"; he has none notable (cannot claim — only the stack can be named); (4) LLM depth is integration-side (LiteLLM APIs, custom GPTs), not solution-building — hedged correctly, but Azure-OpenAI-native candidates will out-depth him here.
- **Cosmetic:** "design documents", "growth mindset", JavaScript depth, C.

### Methodology Transfer Test (top 5 achievements)
| Achievement | How the ISE reviewer sees it |
|---|---|
| SW-7 governed data products for agentic AI | "He already solves the governed-grounding problem our enterprise-RAG engagements hit — permission-aware, metadata-managed data feeding LLM retrieval." ✓ natural |
| BS-1 ML inference into 24/7 fab | "Deploying and operating AI in production under constraints harsher than most customer sites — no maintenance windows, yield on the line." ✓ natural — flagship |
| SW-2 Component Owner, on-call ETL | "That's our DRI model: accountable, on call, restores service." ✓ natural |
| BS-3 Spotfire platform co-ownership + training users | "Customer-facing platform engineering — internal customers, but the code-with muscle is there." ✓ natural |
| SW-1 AWS migration of his domains' ETL | "Cloud migration delivery — wrong cloud for us, right shape of work; concepts transfer." ✓ with the honest-AWS framing doing the bridging |

All five transfer sentences write naturally — the reframing has landed. No bullet requires the reader to do the translation themselves.

### Competitive Landscape
- **Obvious fit:** Azure-native senior SWE from a consultancy (Avanade/Accenture) or ex-product-team Microsoft engineer with OSS visibility and Azure OpenAI project work.
- **Dennis's edge:** genuine enterprise data-platform depth (the *foundation* layer of every AI engagement), regulated-industry scars (telecom, insurance, semiconductor), production ML under 24/7 constraints, native German for DACH accounts, Bern-resident EU citizen (zero visa/relocation friction), cross-industry ramp record that mirrors ISE's engagement model.
- **Their edge:** Azure service fluency, LLM solution-building portfolios, public OSS contributions.

---

## Part 2: Five-Perspective Read-Through

### ATS Robot (keyword scan)
| # | Keyword | Match |
|---|---|---|
| 1 | Python | YES verbatim (bold, multiple) |
| 2 | Java | YES verbatim |
| 3 | C# | YES verbatim |
| 4 | C++ | YES verbatim |
| 5 | JavaScript | YES verbatim |
| 6 | AI systems in production | SEMANTIC (ML inference deployment; Production AI Foundations) |
| 7 | LLM | YES verbatim |
| 8 | RAG | YES verbatim |
| 9 | prompt engineering | YES verbatim |
| 10 | model evaluation | **NO** |
| 11 | data quality | YES verbatim |
| 12 | performance monitoring | YES verbatim |
| 13 | cloud | YES verbatim |
| 14 | modern software engineering practices | SEMANTIC (CI/CD, code review, TDD) |
| 15 | DRI / on-call / restore | SEMANTIC-STRONG (on-call SLA, incident response and restoration) |
| 16 | observability | YES verbatim (group name) |
| 17 | open source | **NO** |
| 18 | cross-functional | **NO** |
| 19 | stakeholders | YES verbatim |
| 20 | German | YES verbatim |

**Match rate: 17/20 = 85% — PASS.** Top 3 truthfully addable: model evaluation (RQ line), open source (3× in JD), cross-functional (pod descriptor).

### Recruiter Glance (10 seconds)
**Verdict: FORWARD (~80%).** "Staff Data & AI Engineer" at Swisscom + tagline with Python/Java/AWS/Kubernetes + "Production AI Foundations & Co-Engineering" + Bern/German-citizen/Zürich-ready/travel-ready line answers every logistics question a Swiss recruiter has before they've read a bullet. Summary's first two lines land data products + AWS + agentic AI. Non-technical reader still understands what he does. Only soft spot: "Co-Engineering" is ISE-insider vocabulary that a generic recruiter may skim past — but this req's recruiter knows the org's own word.

### HR Screen (30 seconds)
**Verdict: PHONE SCREEN (~75%).** Degree ✓ (M.Eng., thesis 1.0), 4+ years coding ✓ (11+), languages of the JD list visibly present ✓, AI-in-production familiarity visible in both summary and a dedicated "AI & ML in Production" skills group ✓, German ✓, travel ✓. First bullet per position is the strongest JD-relevant one in each block (SW-7 grounding for agentic AI; BS-1 production ML; FC-2 applied ML; VZ-1 scale; GN-1 ownership/training). The one RQ term a checklist-reader can't tick verbatim: "model evaluation."

### Hiring Manager (2 minutes)
**Verdict: INTERVIEW (~60%).**
**Top 3 observations:**
1. The Bosch bullet is the credibility anchor — "deploying and operating AI systems in production" answered with a 24/7 wafer fab, which is harsher than most customer environments they scope.
2. The Swisscom story maps onto their current engagement portfolio (governed data → enterprise RAG grounding), and the claims are scope-honest ("within Swisscom's company-wide Data Mesh", "my domains' ETL stack") — no solo-hero smell.
3. The gap they'll price in: AWS not Azure, and LLM work that is integration-grade rather than solution-building-grade. The honest framing earns trust but doesn't erase the gap against Azure-native applicants.
**Predicted first interview question:** "Your cloud depth is AWS — walk me through how you'd ramp onto Azure in the first weeks of a customer engagement."

### Technical Reviewer (10 minutes)
**Truthfulness: PASS — all claims verified.**
| Claim | Verified? | Source |
|---|---|---|
| Governed data products within company-wide Data Mesh (scoped) | ✓ | memory: swisscom_datamesh_ownership — verb/object scoped exactly as mandated |
| ML inference into 24/7 fab, 300mm lines | ✓ | BS-1, bundle flagship |
| Spotfire platform co-owned, Application Owner | ✓ | memory: taf_2022_spotfire ("co-owned" hedge correct) |
| TAF 2022 co-presentation (CL only) | ✓ | memory-verified; correctly reserved for CL |
| LiteLLM, custom GPTs, Copilot, Kiro; **no LangChain** | ✓ | config ban respected — verified toolchain only |
| "Contributed" ML/NLP to ARTUS | ✓ | hedged verb as mandated |
| Component Owner on-call, incident response | ✓ | his components only — scope-disciplined |
| 11+ years | ✓ | May 2015 → present = 11.2 yrs (accurate; session file's "12+" was the estimate, resume is the corrected figure) |
| AWS SAA active to Sep 2027; no Azure claim anywhere | ✓ | config + honest-AWS strategy |
| Security Champion excluded | ✓ | forced exclusion honored (JD has no security gate) |
| Capgemini absent, Generali = Hamburg, education dates/overlap | ✓ | memory corrections all honored |

**Consistency: 1 minor precision note.** Summary clause "run Python services on Kubernetes under on-call SLA" fuses two true facts (SW-3: operates Python apps on K8s; SW-2: on-call SLA as Component Owner for the Fulfillment ETL). A probing interviewer asking "what's the SLA on your K8s services?" gets an answer about the ETL pipelines instead. Defensible but worth a comma's worth of precision — Tier 2.
**Over-saturation:** none. Highest-frequency terms: Python (~10 incl. skills — acceptable as the JD's lead language), AWS (~7), Kubernetes (~5). No term past the concern threshold in bullet prose.

---

## Part 3: Eight-Dimension Scoring

| Dimension | Score | Weight | Weighted | Notes |
|---|---|---|---|---|
| ATS Keywords | 8.0/10 | 15% | 1.20 | 85% match; misses "model evaluation" (RQ line), "open source" (3×), "cross-functional" — all truthfully addable |
| Summary | 8.5/10 | 10% | 0.85 | Strong bridge (data products → AI foundation → customer-embedded); "LLM" absent from recruiter window; one precision nit |
| Skills Section | 8.5/10 | 10% | 0.85 | Group names are domain-perfect ("AI & ML in Production", "Observability & Engineering Quality"); verified GenAI toolchain, zero fabrications |
| Bullet Quality | 8.5/10 | 25% | 2.13 | 18 bullets, all char-clean, scope-disciplined, every top bullet passes the transfer test; light quantification is a KB limit, not a craft failure |
| Publications / Credentials | 8.0/10 | 10% | 0.80 | No pubs (industry resume) — certs carry it: AWS SAA active, Udacity 2026 (fresh), IBM AI Engineering (props up model-eval claim), thesis 1.0 |
| Narrative Coherence | 9.0/10 | 15% | 1.35 | The cross-industry arc IS the ISE pitch; headline → summary → position headers tell one uninterrupted story |
| Page Fill & Visual | 7.0/10 | 5% | 0.35 | 2pp clean, no orphans; page 2 ends ~2/3 down — exceeds ≤3-line rule, documented as unavoidable without KB-unsupported padding (sent Google baseline had the same and cleared screens) |
| Credibility Signals | 8.0/10 | 10% | 0.80 | CNN/BBC/Al Jazeera, 300mm fab, Component/Application Owner titles, AWS cert; no OSS/pubs, few hard metrics |
| **Total** | | **100%** | **83.3** | |

**Score: 83.3/100** — Strong (80–84 band): 1–2 targeted improvements push toward ceiling.

---

## Part 4: Interview Likelihood

| Reader | Probability | Key Factor |
|--------|------------|------------|
| ATS | 90% PASS | 85% keyword rate; all languages verbatim |
| Recruiter (10s) | 80% forward | Staff title + Zürich-ready/German/travel logistics line |
| HR (30s) | 75% phone screen | Every RQ checkable except "model evaluation" verbatim |
| Hiring Manager (2m) | 60% interview | Bosch 24/7 production-AI anchor + zero-friction Swiss hire vs. Azure-native competition |
| Technical Panel (10m) | 70% yes | Claims will survive probing; model-eval/LLM-building depth is the soft spot |

**Ceiling analysis:**
| Scenario | Score |
|---|---|
| Current | 83.3 |
| + Tier 1 applied | ~85.3 |
| Theoretical max (this candidate + this JD) | ~86.5 |
| Hard ceiling (structural: Azure, OSS contributions, LLM solution-building) | ~88 |
| What would close the rest | An Azure cert (AZ-104/AZ-305) or one public OSS contribution — not resume-editable today |

---

## Part 5: Actionable Improvements

### Tier 1 (HIGH — do these)
1. **Add "model evaluation" to the AI & ML in Production skills line.** (+1.0)
   - Current: `\skilldash{\textbf{ML} inference deployment (Docker/Kubernetes), MLOps, data-quality \& performance monitoring}`
   - Proposed: `\skilldash{\textbf{ML} inference deployment (Docker/Kubernetes), MLOps, model evaluation, data-quality \& performance monitoring}`
   - Why: completes the JD's RQ triple ("model evaluation, data quality, and performance monitoring") verbatim. Truthful at the JD's own bar — it asks for *understanding*, backed by IBM AI Engineering Specialization (model building/eval coursework), operating an image classifier in production at Bosch, and thesis NN work. ~108 rendered chars — stays 1 line.
2. **Name the open-source stack.** (+1.0)
   - Current (Software & Data Engineering, line 3): `\skilldash{\textbf{Apache Kafka}, \textbf{Apache Airflow}, batch \& streaming ingestion, \textbf{PySpark} / Spark, Hadoop/Impala}`
   - Proposed: `\skilldash{Open-source data stack: \textbf{Apache Kafka}, \textbf{Apache Airflow}, \textbf{PySpark} / Spark, Hadoop/Impala; batch \& streaming}`
   - Why: JD says "open source" 3× and "variety of technologies, not just Microsoft" is ISE identity. He cannot claim contributions (session tripwire — correctly not claimed), but the stack he runs *is* open source; naming it is a free, honest ATS + culture hit. Verify char count with char_count.py after edit.

### Tier 2 (MEDIUM — optional)
1. **"cross-functional" in the summary.** "I work embedded with customer and stakeholder teams" → "I work embedded in cross-functional customer and stakeholder teams" (+0.5; check summary stays 5 lines).
2. **Summary precision on on-call clause.** "…and run Python services on Kubernetes under on-call SLA" → "…run Python services on Kubernetes, and hold on-call responsibility as Component Owner" (+0.4 probe-proofing; the current phrasing binds the SLA to the K8s services, which the KB doesn't strictly support).
3. **"LLM" into the summary's agentic clause.** "data foundation for agentic AI workloads" → "data foundation for agentic AI and LLM workloads" (+0.3 recruiter-window reinforcement).
4. **Conditional:** if `experience_bosch.md` or `experience_swisscom.md` holds a verified metric (throughput, latency, cost, team count) for BS-1 or SW-1, add it (+0.5). Do NOT invent one.

### Tier 3 (COSMETIC — skip)
1. Page-2 bottom whitespace (~1/3 page) — accepted limitation; padding would violate anti-fabrication.
2. "design documents" / "growth mindset" keyword stuffing — reads as pandering.
3. CL section "Certifications & Awards" contains only certifications — standard template heading, leave.

**Verdict: Apply Tier 1 (both edits, ~5 minutes via /edit-resume). Tier 2 items 1–3 are cheap and worth taking in the same pass. Tier 3 skip.**

---

## Part 6: Interview Bridge Points

| Resume Topic | ISE Equivalent | Opening Line |
|---|---|---|
| SW-7 governed data products for agentic AI | Enterprise RAG grounding / permission-aware retrieval | "The SharePoint-permissions-to-AI-Search problem your team blogged about is the problem I work daily: making governed, access-controlled data queryable by LLM workloads without breaking its contracts." |
| BS-1 ML inference into 24/7 fab | "Deploying and operating AI systems in production" | "A wafer fab has no maintenance windows — a bad deployment costs yield. That constraint taught me the deployment discipline I'd bring to customer production systems." |
| SW-2 Component Owner + on-call | DRI model | "Component Owner at Swisscom is your DRI: I monitor, I get paged, I restore, and I write the runbook so the next restore is faster." |
| BS-3 Spotfire platform + TAF 2022 talk | Code-with / customer-facing engineering | "My customers were internal — fab engineers — but the loop was the same: co-own the platform, extend it in their stack, train them, present it publicly." |
| AWS → Azure | Cloud-agnostic engagement readiness | "I'm certified on AWS at the architecture level; the primitives map — my first week on an Azure engagement is vocabulary, not concepts. ISE's own playbook says meet customers in their stack." |
| LiteLLM + custom GPTs | LLM integration / prompt engineering | "I've built LLM API integrations through LiteLLM and grounded custom GPTs in domain knowledge — the unglamorous 80% of enterprise LLM work is data grounding, and that's my home turf." |
| Cross-industry ramp (5 industries, 3 countries + Shanghai) | ISE engagement model | "Every 1–3 years my job has been: walk into an unfamiliar enterprise, learn its domain, ship production code. ISE just compresses that cycle." |
| Model evaluation (prep for the probe) | Eval understanding | "Operating a defect classifier means watching its precision drift against fab ground truth; on the data side I enforce quality contracts — I know eval from the operational end, and IBM's AI Engineering curriculum covered the formal end." |

---

## Part 7: Cover Letter Critique

### 6A. Anti-Pattern Checklist — PASS 8/8
- ✓ No generic opener (opens with the Engineering Fundamentals Playbook)
- ✓ No CV rehash — adds context (yield framing, "one company at a time")
- ✓ Names specifics: Playbook, ISE SharePoint→AI Search RAG post, USD 400M Swiss datacenter
- ✓ Clear "why THIS role": "doing that across many enterprises instead of inside one" (P1)
- ✓ Strongest qualification early (P1 working-style match, P2 opens with the Swisscom AI-data foundation)
- ✓ No apologetic gap language — AWS handled positively ("the practices transfer, and I learn platforms quickly")
- ✓ Active close ("glad to talk through where your current engagements need this profile")
- ✓ Credentials woven into body (SAA in P4, TAF in P3)

### 6B. Tailoring Signals — PASS 5/5
Playbook + ISE blog post + datacenter investment; JD terms supplementing resume (enterprise RAG, AI Search, side by side, observability as fundamentals, data residency); culture reference (code-with, languages-and-frameworks); specific candidate-method↔need connection (governed data ↔ permission-propagation RAG); institution type correctly industry, tone matches.

### 6C. Industry Checks — PASS
Business value present ("a failed deployment costs yield"; data residency for regulated customers); no academia-exit framing needed; jargon technical but HM-readable per plan (deliberate choice, first reader is likely the ISE team, not central HR).

### 6D. CL ATS Keywords — PASS
Supplements resume with: enterprise RAG, AI Search, LLM solutions, production code, observability, on-call, data residency, German, travel — 8+ of 10 high-priority terms present across the package.

### 6E. Structural — PASS
299 words / 4 paragraphs / 1 page ✓ (industry 250–300). Every claim traceable to a resume bullet or verified memory (TAF 2022 = memory-verified; deliberately CL-only as a deepener of BS-3). Quantified: USD 400M, 24/7, 25%-travel-implied "ready for the travel", req number. Sentence lengths vary (8-word closer to 30-word openers); one contraction ("I'm"); human details land ("costs yield", "one company at a time"). 0 em-dashes.

### 6F. Package Cohesion — PASS
Resume stands alone ✓ (interview-earning without CL). CL deepens rather than introduces ✓. No contradictions in dates, titles, claims ✓. Not a prose restatement ✓. Page budget 2+1=3 ✓.

**Hook verification (step 8b):** all three named artifacts verified live 2026-07-03 with URLs logged in the session file (Playbook; devblogs.microsoft.com/ise/sharepoint-doc-level-access; USD 400M announcement 2025-06-02). No factual errors found.

---

## Part 8: Post-Generation Verification

### Mechanical
- [x] Char limits: 18/18 variable bullets within max (189–210, max 218, zero OVER)
- [x] Orphan check: all 2L bullets fill line 2 ≥70% (visual PDF check)
- [x] Page fill: **EXCEPTION** — page 2 ends ~2/3 down (> 3-line rule); documented accepted tradeoff, matches sent-and-cleared Google baseline
- [x] Bullet ordering matches approved Phase 1 plan (incl. both authorized fillers FC-3, GN-2)

### Content
- [x] ATS ≥70% (85%)
- [x] Provenance flags respected (no false publication/award claims; Security Champion excluded; funding n/a)
- [x] No forbidden terms (LangChain absent; no LOC/test counts; no code-folder names)
- [x] No inflation (hedged: "Contributed" ARTUS, "Co-owned" Spotfire; scoped: "within company-wide Data Mesh", "my domains' ETL stack")
- [x] Publications n/a (industry resume)
- [x] CL claims traceable to resume/memory
- [x] AI fingerprint: banned words CLEAN, 0 rendered em-dashes (both docs), no vague -ing bullet endings, no generic CL opener, sentence variety OK

### Structural
- [x] "Microsoft", "Zürich", "Swisscom", "Fraunhofer" spelled correctly throughout
- [x] Both .tex files compile standalone (MiKTeX, 2pp + 1pp)
- [x] Date format consistent (Mon YYYY -- Mon YYYY); education overlap preserved as mandated
- [x] Email dennis@thiessen.io in both ✓ (config match)
- [x] Page counts: resume 2, CL 1 ✓

**Failures escalated to Tier 1: none.** (Page-fill exception documented, not escalated — padding would require KB-unsupported content.)

---

*End of critique — Pass 1, 2026-07-03. Lens persists for any re-critique.*