Back to Jobs
Recently postedFull-timeOther

Member of Technical Staff - Applied Scientist I [Reinforcement Learning and Simulation Environments]

Ntt Data AivistaSan Francisco Bay Area🇺🇸United States

Quick overview

Salary$300k - $400k/yr
Employment typeFull-time
Work modeHybrid
LocationSan Francisco Bay Area, United States
PythonMachine Learning

Job description

Location: San Francisco Bay Area | Bellevue, Washington | Hybrid

About AIVista

NTT DATA AIVista, Inc., a wholly owned subsidiary of NTT DATA, is based in Silicon Valley. We develop AI products for enterprises operating in complex regulatory environments. We work with NTT DATA Inc., NTT DATA Japan, and NTT DOCOMO to deploy systems for NTT clients. Our clients combine AIVista's product AI expertise with NTT DATA's industry knowledge and systems-integration experience.

The Role

The Applied Scientist will build the learning environments that let AIVista train and evaluate agents before they act in real regulated workflows. You will convert business processes, rules, governed actions, systems of record, documents, and event logs into synthetic, versioned simulations with explicit state, observations, actions, constraints, human-review points, and terminal outcomes.

This role blends reinforcement learning research with environment engineering. You will design reward and cost functions, generate curricula and counterfactual cases, validate simulator fidelity against real operations, and develop offline, safe, constrained, and multi-agent learning methods. The objective is not a demonstration that optimizes a proxy. It is a training and evaluation system that exposes reward hacking, respects governed action boundaries, and predicts whether a policy improves process outcomes without creating unacceptable risk.

Responsibilities

  • Build synthetic environments and workflow digital twins for regulated enterprise processes, including stochastic events, exceptions, human decisions, delays, and external-system behavior.
  • Translate process models, rule sets, action contracts, event logs, and operational policies into Markov decision processes, partially observable models, stochastic games, and discrete-event simulations.
  • Define state, observation, action, termination, reward, cost, and constraint interfaces that are faithful to the underlying workflow and reusable across training and evaluation.
  • Design reward functions and verifiers, then test them for specification gaming, shortcut learning, reward tampering, and unintended incentives using advers and hidden evaluation cases.
  • Validate environment fidelity against production traces, process-mining results, subject-matter expert review, and counterfactual scenarios; quantify where the simulation does not represent the real process.
  • Generate synthetic trajectories, rare-event scenarios, and curriculum stages that improve coverage while preserving provenance and reproducibility.
  • Develop algorithms across offline, safe, constrained, model-based, hierarchical, and multi-agent reinforcement learning, as well as imitation and preference-based learning where appropriate.
  • Integrate deterministic rule checks, governed actions, safety constraints, and human-review policies into training and rollout evaluation.
  • Create benchmarks for sample efficiency, generalization, constraint violations, robustness, calibration, latency, cost, and process-level outcomes.
  • Build scalable environment runners, trajectory stores, experiment infrastructure, and reproducible evaluation pipelines with engineering partners.
  • Move successful research into production decision-support and agent-training systems, with monitoring for distribution shift and policy regressions.
  • Lead publications, patents, open-source contributions, and external research partnerships when they advance AIVista's product and science objectives.

Qualifications

Basic

  • PhD in computer science, machine learning, artificial intelligence, operations research, industrial engineering, applied mathematics, or a related field, or equivalent industry impact.
  • At least 5 years of experience in artificial intelligence or machine learning, with significant experience applying science and research in production environments.
  • A proven record of impactful research demonstrated through publications, patents, open-source systems, benchmarks, or shipped AI products.
  • Deep experience designing reinforcement-learning environments, simulators, reward functions, or sequential-decision systems, not only training policies in existing benchmarks.
  • Strong foundations in reinforcement learning, stochastic processes, optimization, statistical methods, causal reasoning, and experimental design.
  • Proficiency in Python and a modern machine-learning framework, with experience building reliable research software that engineering teams can transition to production.
  • Ability to work with domain experts and translate operational processes, controls, and failure modes into precise learning and evaluation artifacts.

Preferred

  • Experience with offline, safe or constrained, multi-agent, model-based, hierarchical, or risk-sensitive reinforcement learning.
  • Experience designing and implementing custom single-agent or multi-agent learning environments, scalable environment runners, and discrete-event process simulations.
  • Experience with process mining, Business Process Model and Notation, workflow engines, digital twins, operations research, scheduling, or queueing systems.
  • Experience with language-model agents, reinforcement learning with verifiable rewards, verifier-guided training, tool-use agents, or long-horizon agent benchmarks.
  • Experience detecting reward hacking, specification gaming, data leakage, simulator exploitation, or overfitting to public evaluation cases.
  • Experience with synthetic data, counterfactual generation, domain randomization, curriculum design, rare-event simulation, or sim-to-real evaluation.
  • Experience using ontologies, rule systems, action schemas, event logs, or system-of-record interfaces to define environment semantics.
  • Experience deploying learning or decision systems in regulated or high-stakes domains such as insurance, financial services, healthcare, or public-sector operations.

Benefits

  • Medical, dental, and vision insurance
  • 401(k) plan
  • Significant company Health Savings Account contribution
  • Paid holidays and flexible paid time off

Compensation

The estimated base salary range for this role is $300,000 to $400,000 USD. The salary for the successful applicant will depend on job-related factors such as education, training, work experience, business needs, and market demands. This range may be modified in the future. Total compensation also includes variable pay in the form of an annual target bonus and other cash incentives, with a combined potential of up to 100% of base salary.

Equal Opportunity Employment

NTT DATA AIVista is an equal opportunity employer. We do not discriminate based on race, religion, color, national origin, ancestry, sex, gender identity or expression, sexual orientation, age, disability, genetic information, marital status, military or veteran status, reproductive health decisions, or any other characteristic protected under applicable federal, state, or local law.

Haystack's AI agent can fill in and submit this application for you. How AI auto apply works

Will your resume get past this employer's screening? Check your resume against this job, free

Similar jobs