Avatar for agilesoft
Agilesoft provides enterprise software and web solutions for diverse industries

Research Engineer — Post-Training, Mountain View, CA

Posted: 2 days ago
Job Location
Remote Work Policy

In office

Visa Sponsorship

Not Available

RelocationAllowed
Skills
Python
Optimization
Debugging
Deep Learning
Reinforcement Learning
Model Evaluation
Dpo
performance analysis
PyTorch
Generative Models
Experiments
Jax
Implementing
Optimizing
Supervised Fine-Tuning
SFT
RLHF
GRPO
PPO
Preference Optimization
Ablation Studies
Model Behavior
Ablations
Training Objectives
Evaluation Pipelines
Rollout
Reward Functions
Model Adaptation Techniques
Training Dynamics
Measurable Improvements
Research Ideas
Model Failures
Model Post-Training
Scalable Training
Synthetic Data Strategies
GPU-Based Training Systems
GPU-Based Model-Training Pipelines
Training Diagnostics
Efficient Implementations

About the job

Research Engineer — Post-Training

Location: Mountain View, CA — Downtown HQ

Employment Type: Full-time

Work Model: On-site, 5 days/week

Department: Research

About Granica

Granica is building the efficiency and intelligence layer for enterprise AI, helping organizations run increasingly powerful AI systems with less data, compute, and infrastructure overhead.

Our work spans three connected areas:

  • Crunch — Optimizing how enterprises store and process data.
  • Large Tabular Models — Developing foundation models that learn from structured enterprise data.
  • Myelin — Making long-running AI agents more efficient and capable.

Our technology operates within customers' environments, keeping enterprise data and intelligence under their control.

We've processed hundreds of petabytes of enterprise data for customers including Snap, Pinterest, Lyft, and ShareChat.

Led by Stanford Professor Andrea Montanari, our research team combines foundational machine learning with practical systems engineering to advance enterprise AI.

Role Overview

Training a powerful model is only the beginning. Making it genuinely useful requires improving how it learns from feedback, reasons, generalizes, and performs on real-world tasks.

At Granica, we're developing new approaches to enterprise AI, from foundation models for structured data to stateful AI agents. Post-training plays a critical role in translating emerging model capabilities into reliable, measurable improvements.

We're looking for a Research Engineer — Post-Training with a strong technical foundation, deep hands-on machine learning expertise, and ideally a background in original ML research.

You'll work directly with our research scientists to develop and implement post-training methods, run rigorous experiments, and improve model capabilities through a combination of research insight and exceptional engineering.

This is a hands-on research engineering role for someone who enjoys working at the intersection of new ML methods, large-scale training, and practical experimentation.

What You'll Work On

  • Develop and optimize model post-training methods, including supervised fine-tuning, reinforcement learning, and preference optimization.
  • Design and implement training objectives, reward functions, synthetic data strategies, and model adaptation techniques.
  • Build scalable training, rollout, and evaluation pipelines for large generative models.
  • Design rigorous experiments, run ablations, analyze training dynamics, and investigate model failures to drive measurable improvements.
  • Optimize GPU-based training systems for performance, efficiency, reproducibility, and reliability.
  • Collaborate closely with Prof. Andrea Montanari and Granica's research team to translate novel research ideas into working implementations and advances in model capabilities.

What We're Looking For

  • PhD in Computer Science, Machine Learning, Applied Mathematics, Statistics, or a closely related field (completed or nearing completion), or equivalent industry research and engineering experience.
  • Demonstrated hands-on experience with model post-training, including SFT, RLHF, GRPO, PPO, DPO, or related techniques.
  • Strong foundations in deep learning, reinforcement learning, optimization, and modern generative models.
  • Strong Python programming skills and deep proficiency with PyTorch or JAX.
  • Experience implementing, debugging, and optimizing GPU-based model-training pipelines.
  • Strong experimental judgment, including model evaluation, ablation studies, training diagnostics, and performance analysis.
  • Ability to independently translate research ideas into efficient implementations, investigate unexpected model behavior, and iterate toward measurable improvements.

Bonus

  • Strong academic or industry ML research background, including original contributions to model architectures, training methods, or optimization algorithms.
  • Publications at leading machine learning conferences such as NeurIPS, ICML, or ICLR.
  • Experience with distributed training, large-scale reinforcement learning, or frameworks such as verl, TRL, DeepSpeed, or Ray.
  • Experience with reward modeling, synthetic training data, agentic RL, or automated evaluation systems.
  • Experience developing or post-training foundation models for structured data, specialized domains, or complex reasoning tasks.

Compensation & Benefits

  • Competitive salary, meaningful equity, and performance bonus for top performers
  • 401(k) with company match, comprehensive health coverage, and unlimited PTO
  • Daily catered meals in our Mountain View office
  • Support for research, publication, and conference participation

At Granica, you'll help build the next generation of enterprise AI—from exabyte-scale data infrastructure, Large Tabular Models (LTMs), and stateful AI agents. Together, we're creating the infrastructure that enables enterprises to own their data, own the intelligence built on it, and scale both efficiently.

Similar Jobs

GVOS  company logo
GVOS
An Edge Cloud for Autonomous Driving
Shef company logo
Shef
Authentic dishes. Homemade. Delivered. Explore who's cooking in your neighborhood
Astranis company logo
Astranis
Building next-generation internet satellites to get the world online
Archesys company logo
Archesys
Improving the government services that impact everyday lives
Nextdoor company logo
Nextdoor
Nextdoor is the private social network for your neighborhood
Svix (YC W21) company logo
Svix (YC W21)
Enterprise ready webhooks service. Open source and cloud hosted