Research Scientist, LLM Evaluation & Post-Training

OneForma; United States; Hybrid; USD 140,000 - 210,000; Full time

Job summary

Centific AI Research seeks an Applied RL Engineer to design RL environments simulating enterprise workflows and train intelligent agents. You’ll bridge RL research with production systems, delivering measurable improvements in AI agent performance and safe, compliant deployments.

Responsibilities include building digital twins, post-training LLM agents with RLHF and PPO, and creating end-to-end training data pipelines. Strong Python & Gymnasium experience required.

Qualifications

• Deep RL expertise: 3+ years hands-on experience with environment design, reward engineering, policy optimization.

Responsibilities

• Design and build custom RL environments (digital twins) simulating enterprise workflows: document processing, compliance, onboarding, support automation

Skills

Deep RL

LLM post-training

Production software

Agentic AI

Python

Tools

Gymnasium

RLlib

Stable Baselines

PyTorch

JAX

TensorFlow

Job description

About Centific

Centific is a frontier AI data foundry that curates diverse, high-quality data, using our purpose-built technology platforms to empower the Magnificent Seven and our enterprise clients with safe, scalable AI deployment. Our team includes more than 150 PhDs and data scientists, along with more than 4,000 AI practitioners and engineers. We harness the power of an integrated solution ecosystem-comprising industry-leading partnerships and 1.8 million vertical domain experts in more than 230 markets-to create contextual, multilingual, pre-trained datasets; fine-tuned, industry-specific LLMs; and RAG pipelines supported by vector databases. Our zero-distance innovation™ solutions for GenAI can reduce GenAI costs by up to 80% and bring solutions to market 50% faster. Our mission is to bridge the gap between AI creators and industry leaders by bringing best practices in GenAI to unicorn innovators and enterprise customers. We aim to help these organizations unlock significant business value by deploying GenAI at scale, helping to ensure they stay at the forefront of technological advancement and maintain a competitive edge in their respective markets.

About Job Role

Senior Applied Reinforcement Learning Engineer Location: Palo Alto, CA or Seattle, WA (Hybrid/Remote)

About the Team

Centific AI Research advances foundational AI models and applications through reinforcement learning, alignment, and human-centered intelligence.

Our mission is to transform data, signals, and human insight into next-generation intelligent systems that redefine enterprise intelligence.

We're building a governed RL environment platform that enables enterprises to safely iterate and improve AI agent workflows through simulation-based learning, bridging human-labeled signal creation with automated RL training for high-stakes operations.

Role Overview

As an Applied RL Engineer, you will design and build RL environments that simulate complex enterprise workflows and train intelligent agents within them. You'll work at the intersection of RL research and production systems, translating customer requirements into bespoke simulation environments and post-training pipelines that deliver measurable improvements to AI agent performance. This role requires deep expertise in both classical RL methodologies and modern LLM-based agent architectures. You'll shape our product direction and help make RL accessible to enterprise customers who need safe, compliant ways to improve their AI systems.

Core RL Competencies

• Foundational RL MDPs & value methods: State/action spaces, Q-learning, DQN, Double DQN, Dueling DQN
• Policy gradient methods: REINFORCE, Actor-Critic, A2C/A3C, variance reduction
• Advanced optimization: PPO, TRPO, SAC, trust regions, entropy regularization
• TD learning: TD(0), TD(λ), eligibility traces, bootstrapping methods
• LLM Alignment & Post-Training RLHF pipelines: Reward model training, preference learning, human feedback integration
• Direct optimization: DPO, IPO, KTO, offline preference optimization
• Group-based methods: GRPO, RLOO, sample-efficient policy improvement
• Reward modeling: Bradley-Terry models, reward hacking mitigation, KL constraints
• Environment Design Gymnasium/OpenAI Gym: Custom environments, observation/action spaces, wrapper patterns
• Reward engineering: Sparse vs. dense rewards, potential-based shaping, intrinsic motivation
• Verifier design: Programmatic reward functions, outcome verification, ground-truth evaluation
• Simulation: Sim-to-real transfer, domain randomization, multi-agent dynamics
• Advanced Techniques Offline RL: CQL, BCQ, IQL for learning from fixed datasets without environment interaction
• Model-based RL: World models, Dreamer, MuZero, learned dynamics
• Hierarchical RL: Options framework, goal-conditioned policies, temporal abstraction
• Imitation & exploration: Behavioral cloning, GAIL, curiosity-driven exploration, UCB

Key Responsibilities

• Design and build custom RL environments (digital twins) simulating enterprise workflows: document processing, compliance, onboarding, support automation
• Post-train LLM-based agents on domain-specific tasks using PPO, GRPO, DPO, and RLHF
• Build end-to-end pipelines converting human-labeled traces into RL training data
• Architect multi-step reasoning agents with tool-calling and closed learning loops
• Design reward functions, verifiers, and validation frameworks for pre-deployment testing
• Translate cutting-edge RL research into production systems; contribute to publications

Required Qualifications

• Deep RL expertise: 3+ years hands-on experience with environment design, reward engineering, policy optimization
• LLM post-training: Experience fine-tuning LLMs using RLHF, DPO, PPO, or similar
• Production skills: Software engineering beyond research with scalable pipelines and training infrastructure
• Agentic AI: Experience with LLM-based agents, tool use, multi-step reasoning
• Technical stack: Strong Python; Gymnasium, RLlib, Stable Baselines; PyTorch/JAX/TensorFlo