Deskripsi Pekerjaan
Shape the Future of Generative AI
At Nexus Intelligence Labs, we are building the foundational models that power the next generation of autonomous agents. We are looking for a Senior Machine Learning Engineer who thrives at the intersection of research and production engineering. You will not just implement papers—you will architect scalable training infrastructure, optimize massive distributed systems, and ship models that redefine industry benchmarks.
We operate with a 'research velocity, production quality' mindset. Our team is small, elite, and well-funded (Series B, $80M+), backed by top-tier VCs including Sequoia and a16z. We value deep technical ownership, radical candor, and a bias for action.
Why This Role Is Different
- Compute Access: Dedicated access to 4,000+ H100 GPUs via strategic cloud partnerships.
- Open Culture: Strong commitment to open-source contributions and publishing at top venues (NeurIPS, ICML, ICLR).
- Impact: Your code directly powers products used by Fortune 500 companies daily.
If you want to solve the hardest problems in ML systems at scale, surrounded by world-class peers, apply now.
Tanggung Jawab
- Design & Implement Large-Scale Training: Architect and optimize distributed training pipelines (FSDP, DeepSpeed, Megatron-LM) for LLMs and multimodal models on 1000+ GPU clusters.
- Model Architecture Innovation: Propose and validate novel architecture improvements (attention mechanisms, MoE routing, quantization-aware training) to improve throughput and model quality.
- MLOps & Infrastructure: Build and maintain robust CI/CD for model training, evaluation, and deployment (Kubernetes, Kubeflow, MLflow, custom internal tooling).
- Performance Profiling & Optimization: Deep-dive into GPU kernel utilization (NCU/NSight), memory fragmentation, and communication overhead (NCCL/RCCL) to maximize MFU (Model FLOPs Utilization).
- Data Engineering at Scale: Collaborate with Data Engineers to curate, filter, and tokenize petabyte-scale datasets; implement efficient data loading pipelines (WebDataset, DALI).
- Evaluation & Alignment: Develop rigorous evaluation frameworks (benchmarks, human preference/RLAIF) to measure model capabilities and safety.
- Technical Leadership: Mentor junior engineers, lead design reviews, and define technical roadmaps for the ML Platform team.
Kualifikasi
- Education: MS/PhD in Computer Science, Machine Learning, or equivalent practical experience (5+ years industry experience).
- Distributed Training Expertise: Proven track record training models >7B parameters on multi-node clusters (PyTorch FSDP, DeepSpeed, JAX/Flax).
- Systems Programming: Strong proficiency in Python and C++/CUDA; ability to write custom kernels (Triton, CUTLASS) for performance-critical paths.
- Transformer Architecture Mastery: Deep understanding of attention variants (FlashAttention, GQA, MLA), RoPE, normalization techniques, and scaling laws.
- Infrastructure Fluency: Hands-on experience with Kubernetes, Slurm, or Ray; familiarity with GPU interconnects (NVLink, InfiniBand, RoCE).
- Open Source / Research Portfolio: Significant contributions to major ML frameworks (PyTorch, Hugging Face, vLLM, TensorRT-LLM) or publications at top-tier conferences.
- Problem Solving: Exceptional ability to debug non-deterministic failures in distributed environments and navigate ambiguity in research-to-production transitions.