5c03e07ded
Simpler story (slide decks unread -> three-bullet status email), proper Claude-chat look with avatars and bubbles on both sides. Canonical trigger prompt switched to English everywhere (homepage, architecture, both adapter generators rebuilt) - V7 consistency green, all checks pass. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01PDKeXvpT6tENSvyQGLV1Uq
2.6 KiB
esco_uri, esco_label, relation, onet_soc, source, confidence, qa_count, generator, generated
| esco_uri | esco_label | relation | onet_soc | source | confidence | qa_count | generator | generated |
|---|---|---|---|---|---|---|---|---|
| http://data.europa.eu/esco/skill/5b26f08b-88bc-45f0-b901-530d7786466b | scientific computing | optional | 15-2051.00 | model-knowledge | high | 0 | gemma3:27b (prompt-designed and spot-checked by Claude) | 2026-07-10 |
scientific computing — Data Scientist
For a Data Scientist, 'scientific computing' isn’t about running simulations of fluid dynamics – it’s the foundation for reliably manipulating data and building models at scale. Daily tasks involve translating statistical or machine learning algorithms (often expressed mathematically) into efficient, executable code. This means going beyond basic scripting; you're frequently implementing custom functions, optimizing performance with vectorization/parallelization, and understanding numerical stability issues that impact model accuracy. Think feature engineering requiring complex transformations, building bespoke loss functions, or efficiently handling large datasets that don’t fit in memory.
Typical tools extend beyond Pandas & Scikit-learn to include NumPy (core for numerical operations), potentially Numba/Cython for performance boosts, and libraries like SciPy for advanced mathematical routines. You'll also encounter specialized packages depending on the domain – e.g., image processing with OpenCV, signal processing with Librosa. Good scientific computing in this context isn’t just getting a result; it’s ensuring that result is reproducible, well-documented (code and derivations), and demonstrably correct through unit testing and validation against known benchmarks.
A common pitfall is treating data science tools as 'black boxes'. Without understanding the underlying numerical methods, you can easily fall prey to issues like overflow errors, precision limitations, or incorrect assumptions about algorithm behavior. Another is inefficient coding – writing Python loops when vectorized NumPy operations would be orders of magnitude faster. Debugging these problems requires a solid grasp of both the statistical/ML concept and how it’s implemented in code.
What 'good' looks like: Code that isn’t just functional but also readable, maintainable, and performant. A Data Scientist proficient in scientific computing can profile their code to identify bottlenecks, choose appropriate data structures for efficiency, and confidently adapt algorithms to new problems or datasets without relying solely on pre-built solutions.
Weitere Anreicherung
Stage-2 source for future practitioner grounding: arXiv cs.LG/stat.ML (ML preprints).