Deskripsi Pekerjaan
Nexus Intelligence Labs is seeking a visionary Senior Machine Learning Engineer to architect and deploy the next generation of large-scale generative AI systems. Based in the heart of San Francisco's Financial District, you will operate at the intersection of research and production, transforming novel architectures into robust, low-latency services that power our enterprise AI platform.
We are a well-funded, stealth-mode startup founded by veterans from Google Brain, OpenAI, and Meta FAIR. We value deep technical rigor, ownership, and the ability to move fast without breaking things. If you have a track record of shipping foundation models or optimizing distributed training clusters, we want to talk to you.
Why Join Us?
- Technical Ownership: End-to-end ownership of the model lifecycle—data curation, architecture design, distributed training, quantization, and serving.
- Compute Budget: Access to dedicated H100/A100 clusters (thousands of GPUs) with zero queue times.
- Compensation: Top-of-market base, significant early-stage equity (ISO), and comprehensive benefits (medical/dental/vision 100% covered, 401k match, $3k annual learning stipend).
- Culture: Low bureaucracy, high trust. No Jira tickets for research spikes. Weekly paper reading groups and quarterly hackweeks.
Tanggung Jawab
- Design, implement, and optimize novel model architectures (Transformers, SSMs, Mixture of Experts) for multi-modal generative tasks.
- Scale distributed training across 512+ GPU clusters using PyTorch FSDP, DeepSpeed, or Megatron-LM; optimize MFU/throughput via kernel fusion and communication overlap.
- Build and maintain a high-performance inference stack (TensorRT-LLM, vLLM, custom CUDA kernels) targeting sub-50ms latency for 70B+ parameter models.
- Curate and process petabyte-scale datasets; design data flywheels incorporating RLHF/RLAIF and synthetic data generation pipelines.
- Collaborate with Product and Research teams to define model capabilities, evaluation benchmarks (MT-Bench, MMLU, custom enterprise evals), and release criteria.
- Mentor junior engineers; establish best practices for MLOps, experiment tracking (Weights & Biases/MLflow), and reproducible CI/CD for ML artifacts.
- Contribute to open-source publications and represent Nexus at top-tier conferences (NeurIPS, ICML, ICLR).
Kualifikasi
- M.S. or Ph.D. in Computer Science, Machine Learning, or equivalent practical experience (8+ years industry experience).
- Expert-level proficiency in Python and PyTorch/JAX; strong systems programming skills in C++/CUDA/Triton for kernel optimization.
- Proven experience training and deploying Large Language Models (LLMs) or Diffusion Models at scale (10B+ params) in a production environment.
- Deep understanding of Transformer architecture internals: attention mechanisms, RoPE/ALiBi, quantization (GPTQ/AWQ/GGUF), and speculative decoding.
- Hands-on expertise with distributed training frameworks (FSDP, DeepSpeed, Megatron) and hardware acceleration (NVIDIA Hopper/Ampere, NCCL/NVLink).
- Strong software engineering fundamentals: Linux internals, Docker/Kubernetes, CI/CD, git, and code review practices.
- Published research at top-tier venues (NeurIPS, ICML, ICLR, ACL, CVPR) or significant open-source contributions to ML infrastructure (e.g., vLLM, Hugging Face, PyTorch Lightning).