Software Engineer, Model Performance Systems

Details

• Performance Benchmarking: Run and automate standard LLM quality benchmarks (GSM8K, MMLU) alongside custom performance suites for specific workloads (e.g., long-context window, KV cache reuse).
• Infrastructure Validation: Create automated acceptance tests for new GPU clusters across x86 and ARM systems, measuring GPU memory bandwidth, networking throughput, and multi-node networking performance.
• Model Dev Experience: Develop and maintain internal GPU-enabled development environments (similar to GitHub Codespaces). You will ensure the team has seamless, high-performance "dev machines" optimized for model experimentation.
• Tool Development: Build and contribute to tools such as InferenceMAX and genai-bench to automate model evaluation and optimization.
• Deep Hardware Profiling: Use PyTorch Profiler and NVIDIA Nsight Systems to collect performance profiles, identify bottlenecks, and debug the NVIDIA compute/networking stack.
• Monitoring & Observability: Develop real-time dashboards and alerts to monitor system health, model startup times, and runtime performance.
• Continuous Integration: Automate performance testing via CI/CD pipelines to catch regressions in model setups before they hit production.
• Optimization Automation: Build tools to find the "Pareto frontier"—identifying the absolute best configuration (latency vs. cost vs. quality) for a given model and workload.
• A Love for Systems & Hardware: You aren’t just interested in the AI; you want to understand GPU memory subsystems, InfiniBand, and how data moves across a cluster.
• An Automation Mindset: You believe that if a task has to be done twice, it should be scripted. You have a passion for stress-testing and fuzzy testing to find the "breaking point" of a system.
• Mathematical Curiosity: A desire to understand the underlying math of Transformers and how it translates into FLOPs and memory requirements.
• Interest in Optimization: You are excited to learn about (or already play with) quantization, speculative decoding, disaggregated serving, and kernel-level optimizations.
• Technical Toolkit: Familiarity with Python, and an eagerness to master the NVIDIA software stack. C++ familiarity is good to have.
• Direct Impact: Your tools will be the gatekeeper for what defines "good" performance for our customers.
• Deep Learning (Literally): You will gain world-class expertise in GPU orchestration and LLM inference that few engineers in the industry possess.
• High Ownership: As a small team of freshers led by experts, you will have the autonomy to build tools from scratch and contribute to open-source projects.
• Competitive compensation, including meaningful equity.
• 100% coverage of medical, dental, and vision insurance for employee and dependents
• Flexible PTO policy including company wide Winter Break (our offices are closed from Christmas Eve to New Year's Day!)
• Paid parental leave
• Fertility and family-building stipend through Carrot
• Company-facilitated 401(k)
• Exposure to a variety of ML startups, offering unparalleled learning and networking opportunities.