RESTORED CLOUD INC. — Restored Cloud builds AI-powered platforms for the U.S. Department of Defense. Our systems generate synthetic enterprise activity — network traffic, user behaviors, documents, and communications — that is realistic enough to be indistinguishable from genuine operations within military cyber training environments.

Location: Hybrid (U.S.) — periodic on-site at government integration facilities

Employment : Full-Time

Clearance : U.S. Citizen required; ability to obtain Secret clearance (TS/SCI preferred)

Compensation : $130,000 – $170,000 + Equity

THE ROLE

You own the intelligence layer of the platform: the adaptive algorithms that make synthetic traffic indistinguishable from real enterprise activity. Our simulated personas don’t follow scripts — they learn and adapt. You build the learning engines, behavioral models, and anti-patterning systems that prevent trainees from identifying generated activity through statistical analysis. Everything runs on CPU in air-gapped networks with no cloud, no GPU, and no internet.

WHAT YOU OWN

• Adaptive learning engine: multi-armed bandit algorithms (UCB1/Thompson sampling) for action selection with reward signals based on detection avoidance and operational metrics

• Persona behavioral models: state machine transition matrices with role-aware branching and scenario phase progression (baseline → perturbed → contested)

• Anti-patterning systems: template rotation, timing jitter, behavioral drift — preventing statistical fingerprinting of generated traffic

• Realism metrics: quantitative scoring of how well synthetic traffic blends with real enterprise baselines

• Drift detection: tracking baseline shifts in authentication rates, file I/O patterns, and network behavior to keep generated activity aligned with current norms

• Model optimization: ONNX Runtime, INT8 quantization, distilled models — everything must run efficiently on CPU-only compute nodes

REQUIRED QUALIFICATIONS

• 3+ years applied ML or statistical modeling in production systems (not just research/training)

• Strong implementation experience with multi-armed bandits (UCB1, Thompson Sampling, contextual bandits)

• Proficiency in Python with NumPy, SciPy, and statistical libraries

• Experience with Markov models, state machines, or reinforcement learning in deployed systems

• Ability to work within strict compute constraints: CPU-only, limited memory, no cloud offload

• U.S. citizenship required

PREFERRED QUALIFICATIONS

• ONNX Runtime and model quantization (INT8/FP16) experience

• Bayesian methods (prior specification, posterior updating, conjugate models)

• Anomaly detection or behavioral analytics background

• Familiarity with MITRE ATT&CK framework and adversary TTPs

• Prior defense or cyber range operations experience

• Deterministic/seeded random systems for reproducible simulations

THIS ROLE IS NOT FOR YOU IF

– Deep learning researchers who need GPUs and large datasets — our models are lightweight and learn locally

– Data scientists focused on dashboards and visualization — this is an engineering role building production systems

– Candidates who can’t articulate the difference between exploration and exploitation — that’s the core of this work