Deskripsi Pekerjaan
Shape the Future of Generative AI
At Nexus Intelligence Labs, we aren't just building models; we are architecting the cognitive infrastructure for the next generation of enterprise software. We are a well-funded, stealth-mode startup founded by veterans from Google DeepMind, OpenAI, and Stanford AI Lab, backed by Sequoia and a16z.
We are looking for a Senior Machine Learning Engineer to lead the design, training, and deployment of large-scale foundation models. You will own the full lifecycle—from data curation and distributed training on massive GPU clusters to low-latency inference optimization for production.
Why This Role is Unique
- Greenfield Architecture: No legacy baggage. Design the MLOps stack from scratch using modern tooling (Ray, Kubernetes, Dagster).
- Compute Access: Dedicated access to 1000+ H100 GPUs via strategic cloud partnerships.
- Research-to-Production Velocity: Publish at top-tier venues (NeurIPS, ICML) while shipping code to millions of users weekly.
- Equity Upside: Generous pre-IPO equity package (top 1% percentile for early employees).
If you want to solve the hardest problems in AI with a world-class team and zero bureaucracy, apply now.
Tanggung Jawab
- Architect & Train LLMs: Design novel transformer architectures and scaling laws for dense and Mixture-of-Experts (MoE) models; manage distributed training runs (FSDP, DeepSpeed, Megatron-LM) across 512+ GPU clusters.
- Data Centric AI: Build petabyte-scale data pipelines for pre-training, instruction tuning, and RLHF/RLAIF; implement automated data quality filtering, deduplication, and curriculum learning strategies.
- Inference Optimization: Deploy models using TensorRT-LLM, vLLM, and custom CUDA kernels; optimize for latency, throughput, and cost (quantization: GPTQ, AWQ, FP8).
- MLOps & Infrastructure: Own the training/inference infrastructure (Kubernetes, Slurm, Ray); build observability dashboards for GPU utilization, MFU, and model drift detection.
- Research Translation: Rapidly prototype ideas from latest papers (e.g., Mamba, RetNet, Ring Attention) and evaluate viability for production workloads.
- Cross-functional Leadership: Partner with Product and Research Scientists to define model capabilities, evaluation benchmarks (MMLU, HumanEval, custom enterprise benchmarks), and release cycles.
- Mentorship: Mentor junior engineers; establish code review standards, testing protocols (pytest, great_expectations), and documentation culture.
Kualifikasi
- Education: MS/PhD in Computer Science, Machine Learning, or equivalent practical experience (5+ years industry ML experience).
- Large Scale Training: Proven track record training models >7B parameters on multi-node clusters; deep expertise in PyTorch, JAX, or TensorFlow internals.
- Systems Expertise: Strong proficiency in CUDA/C++, Triton kernels, NCCL/RCCL communication libraries, and Linux kernel tuning for RDMA/InfiniBand networks.
- MLOps Mastery: Hands-on experience with Kubeflow, MLflow, Weights & Biases, or Dagster; containerization (Docker, Apptainer) and orchestration (K8s, Slurm).
- Generative AI Fluency: Deep understanding of alignment techniques (SFT, DPO, PPO, Constitutional AI), evaluation methodologies, and safety guardrails.
- Open Source Contributions: Significant contributions to major ML frameworks (PyTorch, Hugging Face, vLLM, LangChain, JAX) or published first-author papers at NeurIPS/ICML/ICLR.
- Communication: Ability to distill complex technical trade-offs to non-technical stakeholders; experience writing technical design docs (RFCs) and post-mortems.