Member of Technical Staff, Compute World Models

General Diffusion, Inc.; San Francisco, Northern (CA, KY); Hybrid; USD 180,000 - 240,000; Full time

Job summary

General Diffusion, Inc. seeks a Member of Technical Staff to model world dynamics and forecast workload behavior before real-machine placement decisions. You will own predictors and evidence, collaborating with RL, runtime, and data teams on validation and safe usage of forecasts.

The role emphasizes action-conditioned state prediction, distribution-shift evaluation, and rigorous experimental practice across hardware configurations and workloads.

Qualifications

• Experience with world models, system identification, probabilistic forecasting, learned dynamics or model-based RL including action-conditioned state prediction.
• Experience evaluating models under distribution shift with held-out environments, hardware configurations, workload regimes or time horizons.
• Experience instrumenting or analyzing real compute systems through traces, telemetry, profiler data or performance counters; intuition for latency, throughput and contention.
• Strong experimental engineering skills to turn messy measured behavior into reproducible datasets, training runs, ablations and failure analyses.

Responsibilities

• Develop action-conditioned dynamics models that forecast compute-system behavior across hardware and workload timescales.
• Define target variables and uncertainty estimates for latency, throughput, contention, headroom, with diagnostics for miscalibration and failure modes.
• Design evaluation splits; measure transfer, calibration and degradation beyond aggregate accuracy.
• Use replay and shadow-mode experiments to test predicted consequences before production decisions.
• Work with Measurement & Data Infrastructure to specify trace fields, provenance, and reproducible training/evaluation slices.
• Publish a predictor contract outlining forecasts, uncertainty, assumptions and limits for RL compute environments and placement ensembles.

Skills

World models

System identification

Probabilistic forecasting

Learned dynamics

Model-based reinforcement learning

Action-conditioned prediction

Tools

Execution traces

Telemetry

Profiler data

Performance counters

Job description

Member of Technical Staff, Compute World Models

Model what happens to a workload before a placement decision reaches a real machine.

Status Open

Area Research

Build action-conditioned models that forecast how a workload will behave before a placement reaches a real machine - across latency, throughput, contention, and available headroom. You will own the predictor and the evidence for when it transfers or fails to transfer; RL & Compute Environments owns policy learning, Heterogeneous Runtime & Placement owns execution, Measurement & Data Infrastructure owns evidence plumbing, and Safety & Formal Verification independently determines which actions are allowed.

01 / The work

What you'll work on

• Develop action-conditioned dynamics models that predict compute-system behavior across hardware and workload timescales from observed execution traces and measured outcomes.
• Define target variables and uncertainty estimates for latency, throughput, contention, and headroom, with diagnostics that make miscalibration and failure modes legible to systems partners.
• Design rigorous evaluation splits that withhold architectures, workload mixes, and load regimes; measure transfer, calibration, and degradation rather than relying on aggregate in-distribution accuracy.
• Use replay and shadow-mode experiments to test predicted consequences of candidate placement actions before those predictions inform a production decision.
• Work with Measurement & Data Infrastructure to specify the trace fields, provenance, and reproducible training/evaluation slices the model needs, without taking ownership of the shared data platform.
• Publish a clear predictor contract - forecasts, uncertainty, assumptions, and known limits for RL & Compute Environments and Heterogeneous Runtime & Placement, while leaving policy selection, execution, and action permissioning to their respective owners.

02 / The background

What you bring

• Demonstrated work in world models, system identification, probabilistic forecasting, learned dynamics, or model-based reinforcement learning, including action-conditioned state prediction.
• A record of evaluating models under distribution shift, especially with held-out environments, hardware configurations, workload regimes, or time horizons; able to reason about calibration as well as point error.
• Experience instrumenting or analyzing real compute systems through execution traces, telemetry, profiler data, or performance counters, with practical intuition for latency, throughput, queueing, and contention.
• Strong experimental engineering skills: can turn messy measured behavior into reproducible datasets, training runs, ablations, and failure analyses that other engineers can inspect.
• Judgment about the boundary between predicting system outcomes and choosing or authorizing actions; comfortable supplying evidence to policy, runtime, and safety teams without collapsing those responsibilities.

03 / The evidence

What progress looks like

• A versioned evaluation suite reports forecast error, calibration, uncertainty coverage, and failure slices for latency, throughput, contention, and headroom on held-out architectures and load patterns.
• Replay and shadow-mode results document where the model's action-conditioned forecasts remain reliable, where they break under shift, and the evidence-based conditions for limiting their use.
• An integration contract and reproducible test cases let policy and runtime teams consume forecasts and uncertainty while demonstrating that independent safety checks - model confidence - remain the authority gate.

04 / In the system

Where this role fits

Owns the predictor and its evidence, not the production policy or the independently enforced boundary.