All career routesCareer learning guide
Research engineer
Turn a proposed method into an experiment that can support or refute a claim. Your work connects mathematical assumptions, a simple baseline, independent checks and carefully bounded conclusions.
The work
An ablation improves the average after its failed seeds were omitted. Reconstruct the full paired seed table, distinguish missing/failing runs from completed runs and state which alternative explanations remain.
Preparation
Python, array reasoning and calculus help. Complete gradient checking and controlled experiments before comparing architectures or extending derivative rules.
Your portfolio
A reproducibility report with a baseline, ablation, numerical checks and unresolved limitations.
Skills and evidence
JVPs, VJPs, and custom derivatives
Required route practice
Check derivative linearity, finite differences and higher-order behavior.
Explain the independent check, repeat it under a changed condition and retain the observed result.
Practice project: stages 1, 2.
jaxpr and program inspection
Required route practice
Explain supported primitives and reject unsupported programs.
Explain the independent check, repeat it under a changed condition and retain the observed result.
Practice project: stages 3.
Hand-built HMC/VI diagnostics; optional NumPyro extension
Optional project extension; 4 project preparation lessons lie outside the required route
Compare posterior moments and sampler diagnostics against the exact target; NumPyro remains an optional library extension.
Explain the independent check, repeat it under a changed condition and retain the observed result.
Practice project: stages 2, 3.
MLflow experiment provenance
Optional project extension; 3 project preparation lessons lie outside the required route
Freeze the question, retain failed attempts and interpret complete paired results.
Explain the independent check, repeat it under a changed condition and retain the observed result.
Practice project: stages 1, 2, 3, 4.
Milestones
Build the shared foundation
A reproducible regression fit with checked gradients and explicit random state.
Inspect and verify a differentiable method
A custom derivative with an explained jaxpr
Review your readiness
Explain the assumptions of your method and support a conclusion with correctness checks and an appropriate baseline.
- Which independent calculation checks the mechanism you changed?
- What did you hold fixed in the baseline and ablation, and which seeds changed?
- What alternative explanation remains consistent with your results?
- What information besides a Git commit is needed to reproduce an experiment with changed data?
This is a learning route for practicing the work, not a claim of job readiness. Use the milestone coverage and a reviewer’s feedback to identify what remains to demonstrate. The capability map below separates required route practice from optional extensions. Public checks and self-reported completion are not independent review or hardware qualification.