Location: In-person, San Francisco (Relocation Supported) Type: Full-time | Competitive Pay + Equity Start Date: ASAP
You will design the algorithms that predict system failures before they happen. Cosmic operates at the intersection of AI and physical infrastructure, and this role owns the prediction layer: turning real-time system telemetry into calibrated forecasts that operators can act on.
This is a hard statistical problem. Failure events are rare. The data is noisy, high-dimensional, and non-stationary. The prediction horizons that matter are long (hours to days, not seconds), and the cost of a miss is not a bad trade but a downed system in an environment where downtime has real consequences.
You will:
- Design and implement time series models for long-horizon failure prediction, including survival analysis, hazard modeling, and competing risks approaches
- Build feature engineering pipelines over multivariate, high-frequency system telemetry
- Develop probabilistic forecasting systems with well-calibrated uncertainty estimates, not point predictions
- Own evaluation methodology: define metrics, build backtesting infrastructure, and measure calibration rigor across deployment environments
- Tackle the rare-event prediction problem head on, including approaches to class imbalance, censored data, and distributional shift
- Design algorithms from first principles for problem domains where off-the-shelf models are insufficient
- Build monitoring and alerting systems that track model performance across distributed, volatile compute environments
- Collaborate closely with infrastructure and systems engineers who understand the hardware side of the stack
- Shape the technical direction of the ML function and, over time, grow and lead a small team of engineers in this area
- 4+ years of experience in applied ML, quantitative research, or statistical modeling
- Strong statistical foundations: survival analysis, Bayesian methods, time series modeling, or stochastic processes
- Experience with long-horizon forecasting or event prediction on noisy, real-world data
- Deep fluency in at least one ML framework (PyTorch, JAX, TensorFlow) and the ability to implement custom model architectures, loss functions, and training procedures from scratch
- Rigorous approach to evaluation: you care about calibration, not accuracy on a held-out set. You have built backtesting or cross-validation pipelines for temporal data and understand why naive splits produce misleading results
- Feature engineering intuition for high-dimensional, multivariate time series data
- Comfort working with imbalanced datasets and rare event prediction
- Strong software engineering habits: clean code, version control, reproducible experiments, testable pipelines
- Ability to work independently on open-ended research problems and communicate findings clearly to engineers who are not ML specialists
- Experience with survival models, hazard functions, or competing risks analysis
- Background in reliability engineering, predictive maintenance, or physics-informed ML
- Publications or open-source work in time series forecasting, anomaly detection, or probabilistic prediction
- Prior experience in quantitative finance, trading systems, or actuarial modeling where calibrated prediction under uncertainty was the core deliverable
- Familiarity with system telemetry, infrastructure monitoring, or observability at scale
- Experience leading or mentoring other engineers on a small technical team
Cosmic is a hardware intelligence platform used across critical compute infrastructure. We have access to a proprietary data layer that most organizations generate but nobody has built real prediction on top of. The data exists, the problem is unsolved, and the operators who depend on this infrastructure need predictions they can act on, not dashboards they have to interpret.
If you have spent your career building prediction systems on noisy data and want to apply that skill to a domain where the stakes are physical and the problem is genuinely hard, we want to hear from you.
Email the following to team@cosmiclabs.io:
Subject line: Member of Technical Staff, AI/ML / [Your Name]
In the body:
- Your name
- Why this role and why Cosmic Labs
- What you bring technically
- Soonest available start date
- Education and years of experience
- U.S. work eligibility status
- Availability (dates and times) for a 15-minute phone call over the next seven days
Attachment:
- PDF resume
Applications reviewed on a rolling basis.