Design, build, and evolve systems for distributed model training and batch inference at scale.
Develop workflows for experiment tracking, model evaluation, and registry management.
Optimize training and inference performance across GPU utilization, data loading, and distributed communication.
NOETIK is a frontier therapeutics company using AI to understand human biology and improve oncology outcomes, building foundation models on proprietary multimodal data. It fosters a collaborative, interdisciplinary culture focused on solving hard problems together.
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.
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.
Build and optimize RL infrastructure for large-scale training.
Improve efficiency across compute, memory, networking, and scheduling.
Design low-level optimizations, from kernels to runtime.
Prime Intellect builds the open superintelligence stack, providing full-stack infrastructure for frontier-scale AI post-training and deployment. The company has raised $150M from top investors and is mission-driven, focused on open-source AI and accelerating science.
Build and maintain core infrastructure for Quora's ML platform, ensuring high availability, scalability, and performance.
Build and improve distributed systems serving ML models in production, from Large Recommendation Models to Large Language Models.
Work on GPU model serving, optimizing latency, throughput, and cost to support larger and more capable models.
Quora's mission is to grow the world's collective intelligence through two platforms: Quora for global knowledge sharing and Poe for AI agent collaboration. We are a remote-first company with passionate, collaborative, and high-performing global teams, rooted in transparency and experimentation.
Build the RL post-training stack end-to-end: rollout ingestion, reward computation, policy updates, and returning updated weights to the network.
Invent algorithms for asynchronous, high-latency, partially trusted generation, including staleness tolerance and off-policy corrections.
Ship the first decentralized post-trained models, building evals and taking the release from run to public artifact.
Pluralis Research is developing Protocol Learning, a decentralized approach to training and serving large models on consumer devices. We are a small, deeply technical team backed by Union Square Ventures, with a remote-first culture and team members spread globally.
Lead the effort to make Runpod the fastest and most cost-efficient place for LLM inference, owning performance end to end.
Profile and diagnose performance bottlenecks across the serving stack, from scheduling to kernels, and implement fixes.
Work closely with product and infrastructure teams to shape how inference is offered, turning improvements into production-ready defaults.
Runpod is the AI Developer Cloud, providing a platform for over one million developers to experiment, train, fine-tune, deploy, and scale AI. We're a small, remote-first team that takes ownership seriously, moves fast, and has processed more than 20 billion inference requests.
Build and extend pretraining code to scale experiments into full training runs.
Evaluate model architecture changes and optimize training throughput and resilience.
Run controlled ablations and collaborate with data and evaluation researchers.
Prime Intellect is building an open superintelligence stack, providing the infrastructure and platform for frontier AI training and deployment. The company has raised $150M from top investors and boasts a mission-driven team working to decentralize AI.
Own optimization of the inference stack: throughput, latency, cold starts, and cost per token.
Tune SGLang/vLLM engines, cut TTFT, boost KV cache reuse, and kill cold starts on 1.5 TB checkpoints.
Set technical direction with data, benchmark every change, and scale inference across nodes.
Kimchi is an LLM inference platform serving open models like DeepSeek and GLM from its own GPU fleet via one API line. It's made by CAST AI, a globally distributed team across 34 countries that hit unicorn status in 2026.