Deskripsi Pekerjaan
At Nexus AI Labs, we are architecting the future of generative intelligence. We are seeking a Senior Machine Learning Engineer to lead the development of large-scale foundation models that power our enterprise product suite. This is a high-impact role where you will own the full lifecycle—from data curation and distributed training to optimized inference and production monitoring.
You will join a elite team of researchers and engineers publishing at top-tier venues (NeurIPS, ICML, ICLR). We offer a culture of radical candor, deep technical freedom, and the compute budget to turn ambitious research into shipped product. If you want to push the boundaries of LLMs and multimodal systems alongside world-class peers, this is your next mission.
Tanggung Jawab
- Design, train, and optimize large language models (LLMs) and multimodal architectures using distributed training frameworks (FSDP, DeepSpeed, Megatron-LM).
- Build and maintain scalable data pipelines for petabyte-scale datasets, ensuring data quality, deduplication, and tokenization efficiency.
- Implement advanced optimization techniques (QLoRA, quantization, speculative decoding, kernel fusion) to maximize throughput and minimize latency for production inference.
- Collaborate with Research Scientists to translate novel architectures (Mixture of Experts, State Space Models) from paper to production-grade code.
- Establish and enforce MLOps best practices: experiment tracking, model registry, automated evaluation benchmarks (HELM, MMLU, custom evals), and CI/CD for model deployment.
- Mentor junior engineers and conduct rigorous code reviews focusing on performance, reproducibility, and system reliability.
- Partner with Product and Infrastructure teams to define hardware requirements (GPU/TPU cluster topology, networking, storage) for upcoming training runs.
Kualifikasi
- M.S. or Ph.D. in Computer Science, Machine Learning, or equivalent practical experience (5+ years industry experience building DL systems).
- Expert proficiency in Python and deep learning frameworks (PyTorch preferred, JAX/TensorFlow a plus).
- Proven track record of training or fine-tuning models >7B parameters on multi-node GPU clusters (slurm, kubernetes, Ray).
- Deep understanding of transformer architecture internals, attention mechanisms, and memory optimization (FlashAttention, PagedAttention).
- Strong systems engineering skills: CUDA/Triton kernel writing, NCCL/RCCL communication primitives, and CPU/GPU profiling (Nsight, PySpy).
- Experience with MLOps tooling: MLflow, Weights & Biases, Kubeflow, or internal platform equivalents.
- Publications at top-tier conferences (NeurIPS, ICML, ICLR, ACL) or significant open-source contributions to the LLM ecosystem (vLLM, Hugging Face, DeepSpeed, Axolotl).
- Excellent communication skills; ability to articulate complex technical trade-offs to both technical and non-technical stakeholders.