Design, run, and analyse large-scale RL post-training experiments such as RLHF and RL from verifiable rewards.
Research and develop methods that improve out-of-distribution generalization, including reward design, task diversity, and regularization.
Build held-out evaluations to measure generalization and catch reward hacking, and turn successful research into production training recipes.
Cohere is a leading security-first enterprise AI company building cutting-edge foundation models and end-to-end products for businesses. We are a global team of researchers, engineers, and designers passionate about their craft, headquartered in Toronto with offices worldwide.
Distributed pretraining: Implement and optimize model-parallel training on heterogeneous GPUs under low-bandwidth, high-latency links.
Performance optimization: Apply techniques that reduce communication overhead while maintaining model convergence in challenging network environments.
Elasticity and fault tolerance: Build robust checkpointing and recovery so runs survive node churn as participants join and leave.
Pluralis Research is a decentralized AI research company developing Protocol Learning to train large models on consumer devices across the internet. The team is world-class and deeply technical, backed by tier-1 investors, with a remote-first culture.
Own the inference stack end-to-end, including pipeline-parallel execution, placement and routing, transport, serving engine, and failure handling.
Invent and validate novel algorithms for fast inference on consumer hardware over the public internet, then put them into production.
Serve both RL training rollouts and final model serving in a permissionless, trustless setting.
Pluralis Research works on Protocol Learning: training and serving large models in a fully decentralized way on small consumer-grade devices connected via the internet. They are a world-class, deeply technical team backed by Union Square Ventures and other tier-1 investors.
Design and execute experiments in reasoning model training, post-training alignment, test-time scaling, and systematic model evaluation for crypto-native applications.
Review and synthesize recent academic literature from NeurIPS, ICML, ICLR, and ACL to inform ongoing research directions in reasoning, alignment, and agentic AI.
Implement model variants and evaluation protocols using PyTorch and the Hugging Face ecosystem, tracking experiments with Weights & Biases.
Binance is a leading global blockchain ecosystem behind the world's largest cryptocurrency exchange by trading volume and registered users. Trusted by 300+ million people in 100+ countries, the company has a flat, results-driven culture and focuses on increasing the freedom of money and financial access worldwide.
Build the Learning Gym, a sandboxed environment for training and evaluating AI agents on real Niural workflows with ground-truth reward signals.
Engineer verifiers, run reinforcement learning loops, and establish honest baselines to prove agent improvements generalize.
Write first-author research papers and internal technical reports, and represent the work externally through preprints and talks.
Niural is a global payroll, Employer of Record (EOR), Agent of Record (AOR), and Contractor Management platform. It is a relentlessly ambitious team focused on building foundational internet infrastructure for the AI era.
Own the threat model for decentralized training and inference, enumerating potential attacks and keeping it current.
Design and calibrate statistical verification methods with stated error rates across heterogeneous hardware.
Build and run the verification service in the inference path and take responsibility for its mistakes.
Pluralis Research is building Protocol Learning, a decentralized approach to training and serving large models on consumer-grade devices. They are a small, world-class team of ML researchers backed by Union Square Ventures, with a remote-first culture and an ideological commitment to trustless, sovereign AI.
Build the component layer around our layout synthesis engine, including API contracts, services, and evaluation gates.
Turn model retraining into a one-command job with built-in benchmarks and readable results for the whole team.
Own latency and cost budgets for learned capabilities and partner with infrastructure engineers on MLOps and deployment.
Higharc is a VC-backed startup that is changing how new homes are designed and built using spatial AI and generative floor plan technology. The company is fully remote, has raised over $175M, and values flexibility, collaboration, and asynchronous deep work.
Design high-performance training platform components including orchestration, observability, and performance tuning.
Deliver end-to-end ML pipelines from dataset curation to training, validation, and deployment.
Drive technical direction across ML Platform, Infrastructure, Autonomy, and Safety teams.
Stack is developing revolutionary AI and advanced autonomous systems for safer, more reliable, and efficient operations, focusing on autonomous trucking. The Stack team brings decades of experience in real-world systems and is committed to a culture of inclusion, entrepreneurship, and innovation.