You’ll work hands‑on with instrumentation/test workflows and our AI/ML platform to build a closed‑loop optimization pipeline that raises failure coverage, shortens test time, and increases reproducibility.
Responsibilities
- Model‑guided parameter optimization: Build and evaluate optimization loops (Bayesian optimization, bandits, RL) that select fault‑injection parameters to maximize coverage or detection sensitivity under resource constraints.
- Experiment design & telemetry: Define Design of Experiments/sequential experiment plans, log metrics/parameters/artifacts, and instrument robust telemetry for analysis and replay.
- Adaptive learning: Implement feedback loops so models update with every run, improving the next round of injections (active learning).
- Scalable tuning: Use distributed hyperparameter search to explore large parameter spaces efficiently.
- Reliability metrics: Define and track objective functions (e.g., fault detectability, coverage, time‑to‑fail, false positives/negatives), plus safety/guardrails for destructive tests.
- Impact demo: Deliver a working prototype that can be run by engineers via our platform (scripts + config + dashboards) and present measurable improvements vs. baseline domain‑tuned flows.
Qualifications
- Currently pursuing a MS/PhD in EE/CE/CS or related field.
- Solid Python and ML fundamentals (supervised/unsupervised learning, overfitting, uncertainty).
- Experience with one optimization method (Bayesian optimization, bandits, RL) or hyperparameter tuning at scale.
- Data handling/visualization (NumPy/Pandas/Matplotlib), version control (Git).
Candidates who wish to be considered must be enrolled in a accredited college/university as of September 2026. Applicants who have graduated before September 2026 will not be considered unless they are entering/applying to a MS or PHD program after graduating.
Visa Sponsorship is not available for this position. Candidates who now or at any point in the future require sponsorship for employment visa status (e.g., H-1B Visa status) may not be considered.
