Deskripsi Pekerjaan
Nexus Intelligence Labs is at the forefront of applied AI, building foundational models that redefine how enterprises interact with unstructured data. We are a well-funded, stealth-mode startup backed by top-tier VCs (Sequoia, a16z) founded by ex-Google Brain and Meta FAIR researchers.
We are seeking a Senior Machine Learning Engineer to own the lifecycle of large-scale model training, optimization, and deployment. You will not just implement papers; you will architect the infrastructure that pushes the boundaries of context windows, retrieval-augmented generation (RAG), and multimodal reasoning.
This is a high-autonomy role for a builder who thrives in ambiguity and wants to ship products used by millions. We offer a competitive equity package, comprehensive healthcare, a $3,000 annual learning stipend, and a culture of technical excellence.
Tanggung Jawab
- Design, train, and optimize large language models (LLMs) and multimodal architectures on distributed GPU clusters (1000+ GPUs).
- Architect and maintain high-throughput data pipelines for petabyte-scale datasets, focusing on data quality, deduplication, and curriculum learning.
- Implement advanced training techniques: FlashAttention, FSDP/DeepSpeed, mixed precision, and model parallelism.
- Lead the evaluation framework: build robust benchmarks for reasoning, coding, hallucination detection, and safety alignment (RLHF/RLAIF).
- Collaborate with Product and Research to translate cutting-edge research into production-grade APIs with sub-100ms latency.
- Mentor junior engineers and establish best practices for MLOps, experiment tracking (Weights & Biases/MLflow), and reproducible research.
- Contribute to open-source tooling and publish findings at top-tier conferences (NeurIPS, ICML, ICLR).
Kualifikasi
- M.S. or Ph.D. in Computer Science, Machine Learning, or equivalent practical experience (5+ years industry ML experience).
- Proven track record of training/fine-tuning models >7B parameters (e.g., LLaMA, Mistral, BERT variants, Diffusion models).
- Expert-level proficiency in PyTorch, CUDA kernels, and distributed training frameworks (FSDP, DeepSpeed, Megatron-LM).
- Deep understanding of Transformer architecture internals: attention mechanisms, positional embeddings, quantization (GPTQ/AWQ), and speculative decoding.
- Strong software engineering fundamentals: Python, Rust/Go/C++, Git, CI/CD, Docker, Kubernetes.
- Experience with cloud infrastructure (AWS/GCP/Azure) and cluster management (Slurm, Ray, Kubernetes).
- Publications at top ML venues or significant open-source contributions (GitHub profile required).