feat: rebuild evidence-first application workflow

resume_builder/support/significance_bosch.md

@@ -1,7 +1,7 @@
 # Significance Research: Bosch Semiconductor — Data Analysis Engineer
 
-> Use in cover letters and summaries — NOT in resume bullet text.
-> Particularly valuable for semiconductor industry JDs.
+> Optional semiconductor context only. Reverify external claims before use.
+> Never convert generic fab scale, yield economics or industry trends into Dennis's personal impact.
 
 ---
 
@@ -16,19 +16,19 @@
 - Offline ML analysis (batch — not real-time, misses process drifts)
 - Inline ML inference (real-time, containerized — current best practice)
 
-**Why Dennis's experience matters:** Deploying ML inference into a 24/7 fab is operationally much harder than deploying to a web server. There are no maintenance windows, hardware is constrained, and a model failure affects production throughput. Dennis designed and executed the integration strategy for this environment — a level of MLOps maturity that few data engineers have encountered.
+**Why Dennis's experience matters:** Dennis designed and executed an integration strategy for containerized ML inference in a continuously operating fab environment. Do not add claims about maintenance windows, hardware constraints, throughput impact or rarity unless they are verified for his system.
 
-**Differentiation:** The combination of Docker containerization + Kubernetes orchestration + Ansible automation in a 24/7 constrained environment is a rare and credible production ML deployment signal.
+**Differentiation:** Docker, Kubernetes and Ansible used for production inference integration provide direct deployment evidence. This supports ML-platform/MLOps positioning without implying model-development ownership.
 
 ---
 
 ### Semiconductor Data Domains — Field Context
 
 **Defect Management:**
-Semiconductor defect management involves tracking, classifying, and correlating defects found during inline inspection (optical, SEM) and end-of-line electrical test. Key data challenges: high-dimensional spatial data (wafer maps), multi-step process correlation, and connecting defect signatures to root causes (process excursions, equipment issues). Dennis built data pipelines and ML systems directly in this domain.
+Semiconductor defect management provides the domain context for Dennis's work. His verified contributions cover data services, analytics applications, wafer-map visualizations and ML-inference integration; do not generalize this into ownership of every defect-management pipeline or ML system.
 
 **Semiconductor Parameter Testing:**
-Parametric testing measures electrical characteristics (threshold voltages, leakage currents, resistance) of test structures on each wafer. The data volume is massive — hundreds of parameters across thousands of dies per wafer, across thousands of wafers per day. Data engineering for parametric test requires efficient storage, fast query access, and statistical analysis capabilities. Dennis built data services that fed parametric testing analysis teams.
+Dennis built data services for semiconductor analysis teams in the parameter-testing domain. Generic wafer or data-volume figures must not be presented as the scale of his systems without direct evidence.
 
 **Process Analysis:**
 Process analysis correlates equipment parameters (temperature, pressure, gas flow) with downstream wafer yield and defect outcomes. This is the domain where data engineering meets process engineering — the pipelines must be reliable and the data must be accurate, because process decisions (equipment maintenance, recipe adjustments) depend on it.
@@ -40,13 +40,13 @@ Process analysis correlates equipment parameters (temperature, pressure, gas flo
 ### Field Overview: Data & AI in Semiconductor Manufacturing (2024–2026)
 
 The semiconductor industry is undergoing a major digital transformation driven by:
-1. **Process complexity:** 300mm fabs with 1000+ process steps generate petabytes of data; manual analysis can no longer keep pace
-2. **Yield pressure:** At leading-edge nodes, even 1% yield improvement has enormous economic value — data-driven yield optimization is a strategic priority
+1. **Process complexity:** 300mm semiconductor production creates complex data and operational requirements; do not attach generic petabyte or process-step figures to Dennis's work
+2. **Yield and quality:** Data-driven process and defect analysis matter commercially, but Dennis has no verified personal yield-improvement metric
 3. **AI/ML adoption:** Computer vision for inline inspection, predictive maintenance for equipment, and ML-based process optimization are all actively deployed at tier-1 fabs (TSMC, Intel, Samsung)
-4. **Talent scarcity:** Candidates who combine data engineering depth with semiconductor domain knowledge are extremely rare — most data engineers lack the domain; most process engineers lack the data skills
+4. **Candidate distinction:** Dennis combines production data engineering with direct semiconductor-fab application experience; avoid unsourced scarcity claims
 
 **Target companies for semiconductor JDs:**
 ASML, Infineon, GlobalFoundries, ams OSRAM, Microchip Technology, ON Semiconductor, Renesas, NXP, STMicroelectronics, Bosch (again), TSMC (Europe fabs in Dresden area), Wolfspeed, SiCrystal, Elmos
 
 **CL hook for semiconductor JDs:**
-> "Semiconductor manufacturing analytics is one of the most data-intensive and operationally demanding domains in industry. At Bosch Semiconductor in Dresden, I worked directly in the data domains that matter most — Defect Management, Semiconductor Parameter Testing, and Process Analysis — building the pipelines and analytics platforms that engineers relied on for real-time production decisions. That domain knowledge, combined with my experience deploying ML-based defect classification into a 24/7 fab, is what I'd bring to [Company]."
+> "At Bosch Semiconductor in Dresden, I developed data services and analytics applications for defect-management and process-analysis teams, and integrated containerized ML inference into a 24/7 fab environment. That combination of domain familiarity and production delivery is what I would bring to [Company]."