# Skills Taxonomy — Evidence-First

> Canonical authority: `resume_builder/canonical/claims.json`.
> This file helps select and group skills; it may not promote a skill beyond its canonical evidence level.

## Evidence Levels

| Level | Meaning | Output rule |
|---|---|---|
| Production — current | Used in current professional delivery | May appear plainly when relevant |
| Production — historical | Shipped professionally, but not current | Include with role/date context when recency matters |
| Hands-on — current | Used directly, but without verified production ownership | Use precise verbs such as used, configured or integrated |
| Project / proof of concept | Used in a bounded PoC | Label the PoC; never imply platform-scale operation |
| Certification / coursework | Learned through formal study | Keep in certification context; not professional experience |
| Unverified / never used | No reliable evidence | Do not include |

Do not use Expert/Proficient/Familiar labels in resumes. Evidence and recency are more useful than self-ratings.

## Current Production Core

| Skill | Evidence | Typical use |
|---|---|---|
| Python | Swisscom pipelines/apps; prior Bosch/Vizrt work | Always for data/platform roles |
| SQL | Swisscom and prior data roles | Always for data roles |
| PySpark | Swisscom current work | When distributed processing is relevant |
| Apache Kafka | Swisscom production ingestion | Data/event-driven roles |
| Apache Airflow | Swisscom AWS migration scope | Orchestration/data roles |
| AWS | Swisscom production work; SAA certification | AWS-relevant roles |
| S3, Glue, Athena, Iceberg, Redshift | Swisscom owned-domain migration/data products | Name only relevant services |
| CloudFormation / IaC | Swisscom production provisioning | Say CloudFormation; never substitute Terraform |
| Kubernetes, Docker | Swisscom application delivery; Bosch ML integration | Production platform/MLOps roles |
| GitLab CI/CD | Swisscom delivery | Platform and engineering roles |
| Oracle, Teradata | Swisscom pipelines | Data roles when relevant |

## Historical Production Evidence

| Skill | Evidence | Constraint |
|---|---|---|
| Java | Bosch, Fraunhofer, Generali | Historical; do not imply current daily use |
| C# | Bosch and Fraunhofer | Historical; strong when Spotfire/.NET is relevant |
| C++ | Vizrt distributed backend | Limited historical evidence |
| JavaScript / Express.js | Fraunhofer MISSION | Historical and bounded; TypeScript is unverified |
| Hadoop / Impala | Bosch data services | Historical production context |
| Ansible | Bosch ML integration | Historical production context |
| Jenkins | Fraunhofer and Generali | Historical production context |
| BDD, Selenium, JBehave | Generali | Earlier-career testing evidence |
| TIBCO Spotfire | Bosch co-ownership and C# extensions | Preserve co-ownership |

## ML, AI and Observability

| Skill | Evidence level | Safe framing |
|---|---|---|
| ML inference deployment | Production — historical | Integrated containerized inference into a 24/7 Bosch fab |
| Image classification | Production application context | Worked on inference integration; model-training ownership not verified |
| MLOps | Bounded production evidence | Use only when defined as deployment/operation, not full model lifecycle |
| NLP / speech recognition | Research-project contribution | Contributed components at Fraunhofer; no publication/model ownership |
| ELK, Kafka anomaly detection | Proof of concept | Always retain the PoC label |
| Grafana, Prometheus, Loki | Proof-of-concept/monitoring context | Do not imply enterprise observability ownership |
| LiteLLM | Hands-on — current | LLM API gateway use/integration; no serving-platform ownership |
| Domain-grounded assistants/custom GPTs | Hands-on — current | Configured with curated knowledge; no fine-tuning or formal evaluation |
| Copilot, Kiro | Hands-on — current | AI-assisted engineering tools, not LLM product engineering |

## Certification-Only Signals

| Skill | Evidence |
|---|---|
| TensorFlow / Keras | IBM AI Engineering coursework |
| PyTorch | Coursework/personal evidence only; verify before listing outside certification context |
| Spark ML | Coursework context only unless professional evidence is added |
| AI for Trading / quantitative ML | Udacity/WorldQuant Nanodegree |

## Forbidden Until New Evidence Is Added

- LangChain, LangGraph, LlamaIndex, AutoGen, CrewAI or Semantic Kernel
- Azure, Azure ML, Azure OpenAI or AKS
- GCP, BigQuery, Dataflow or Flume hands-on experience
- Terraform
- Formal model or LLM evaluation
- LLM fine-tuning, red-teaming or model-training ownership
- FastAPI, Flask or Django
- TypeScript
- Petabyte-scale ownership

## Certifications

| Certification | Issuer | Year/status |
|---|---|---|
| AWS Certified Solutions Architect — Associate | AWS | 2024; active to Sep 2027 |
| Data Engineering with AWS Nanodegree | Udacity | 2026 |
| iSAQB CPSA — Foundation | iSAQB | 2016; no expiry |
| ITIL Foundation | PEOPLECERT / AXELOS | 2016; no expiry |
| AI for Trading Nanodegree | Udacity / WorldQuant | 2021 |
| IBM AI Engineering Specialization | IBM / Coursera | Completion year not recorded |

The Swisscom Security Champion assignment is not a certification and does not belong in this table.

## Resume Grouping

Use 4--6 compact lines, selected for the JD. A normal International Tech grouping is:

1. Languages: Python, SQL; selected historical languages only when required.
2. Data: Kafka, Airflow, PySpark, Oracle/Teradata, data products and governance.
3. Cloud/platform: AWS services, CloudFormation, Kubernetes, Docker, GitLab CI/CD.
4. ML/operations: ML inference deployment and bounded observability evidence.
5. Certifications: one line, only the most relevant credentials.

Never add a skill only to mirror a JD. Every listed skill must have a canonical evidence level and an interview-ready example.