Deskripsi Pekerjaan
Shape the Future of Generative AI
At Nexus Intelligence Labs, we don't just deploy models—we architect the foundational infrastructure that powers the next generation of autonomous agents. We are seeking a Senior Machine Learning Engineer to lead the lifecycle of large-scale foundation models, from novel architecture research to high-throughput production deployment.
This is a high-autonomy, high-impact role for an engineer who thrives at the intersection of deep learning theory and distributed systems engineering. You will own critical paths in our training stack, optimize inference latency on custom hardware, and collaborate directly with research scientists to translate theoretical breakthroughs into robust, scalable products used by millions.
Why This Role Is Different
- ✓Compute Access: Unrestricted access to massive H100/A100 clusters via strategic cloud partnerships.
- ✓Open Publishing: Strong encouragement & budget for publishing at NeurIPS, ICML, ICLR.
- ✓Greenfield Architecture: Build the MLOps platform from the ground up—no legacy technical debt.
- ✓Equity Upside: Pre-IPO stock options with quarterly liquidity events.
Tanggung Jawab
- Architect & Scale Training Pipelines: Design and implement distributed training frameworks (FSDP, DeepSpeed, Megatron-LM) for models 70B+ parameters across thousands of GPUs.
- Inference Optimization: Lead quantization (GPTQ, AWQ), kernel fusion (Triton, CUDA), and speculative decoding strategies to achieve sub-50ms latency at scale.
- Data-Centric AI: Build petabyte-scale data curation pipelines (Spark/Ray) for pre-training, instruction tuning, and RLHF/RLAIF alignment.
- MLOps & Observability: Own the CI/CD/CT pipeline for model training, evaluation (HELM, custom benchmarks), and automated rollout/canary deployment.
- Research-to-Production Translation: Partner with Research Scientists to reproduce novel architectures (Mixture of Experts, State Space Models) and harden them for production reliability.
- Technical Leadership: Mentor junior engineers, define coding standards for ML systems, and drive architectural review boards for infrastructure decisions.
Kualifikasi
- Advanced Degree: MS/PhD in Computer Science, Machine Learning, or equivalent practical experience (5+ years industry ML engineering).
- Distributed Training Mastery: Proven track record training models >10B params on multi-node clusters (PyTorch FSDP, JAX, Megatron).
- Systems Fluency: Expert-level Python/C++; deep understanding of GPU architecture (NCCL, CUDA), memory hierarchy, and networking (InfiniBand/RoCE).
- LLM Lifecycle Expertise: End-to-end experience: Tokenization → Pre-training → SFT → Alignment (DPO/PPO) → Serving (vLLM/TGI/TensorRT-LLM).
- Open Source Contributor: Meaningful contributions to major ML frameworks (PyTorch, Hugging Face, JAX, FlashAttention, vLLM).
- Production Rigor: Experience with Kubernetes (Kubeflow/KServe), Terraform, and observability stacks (Prometheus/Grafana/Datadog) for ML workloads.