Deskripsi Pekerjaan
Nexus AI Labs is at the forefront of building next-generation foundation models for enterprise automation. We are a well-funded, stealth-mode startup founded by ex-researchers from top AI labs, backed by tier-1 VCs. We are looking for a Senior Machine Learning Engineer to own the full lifecycle of large-scale model training, optimization, and deployment.
This is a high-impact role where you will architect the training infrastructure for models with 70B+ parameters, pioneer novel RLHF/RLAIF pipelines, and push the boundaries of inference efficiency on custom hardware. You will work alongside a world-class team in a culture that values intellectual honesty, speed, and technical excellence.
Why Join Us?
- Ownership: End-to-end ownership of critical model training and serving paths.
- Compute: Access to massive GPU clusters (H100/A100) with zero queue times.
- Impact: Your work ships directly to Fortune 500 customers transforming their operations.
- Compensation: Top-of-market salary, significant early-stage equity, and comprehensive benefits.
Tanggung Jawab
- Design, implement, and optimize distributed training workflows for Large Language Models (LLMs) across thousands of GPUs using PyTorch FSDP, DeepSpeed, or Megatron-LM.
- Architect and maintain high-throughput data processing pipelines (Petabyte-scale) for pre-training, fine-tuning, and alignment (SFT/DPO/RLHF).
- Lead model optimization efforts: quantization (GPTQ/AWQ), kernel fusion (FlashAttention, Triton), and compilation (TorchInductor/TensorRT-LLM) for low-latency inference.
- Develop rigorous evaluation frameworks (automated + human-in-the-loop) to measure model quality, safety, and hallucination rates.
- Collaborate with Research Scientists to translate novel architectures (MoE, RWKV, Mamba, RAG systems) from paper to production-grade code.
- Build and maintain the MLOps platform: experiment tracking, model registry, CI/CD for ML, and automated regression testing.
- Mentor junior engineers and establish best practices for code quality, reproducibility, and GPU utilization efficiency.
- Stay current with literature; reproduce key papers and assess applicability to our product roadmap.
Kualifikasi
- Education: M.S. or Ph.D. in Computer Science, Machine Learning, or equivalent practical experience.
- Experience: 5+ years of professional ML engineering experience; 2+ years specifically training/serving LLMs (10B+ params) in production environments.
- Framework Mastery: Expert-level proficiency in PyTorch; deep understanding of CUDA, distributed systems (NCCL, MPI), and GPU architecture (Hopper/Ampere).
- Distributed Training: Proven track record scaling training on 100+ GPU clusters (FSDP, ZeRO, Pipeline/ Tensor Parallelism).
- Inference Optimization: Hands-on experience with TensorRT-LLM, vLLM, TGI, or custom Triton kernels for production serving.
- Data & MLOps: Strong software engineering fundamentals (Python, Rust/C++, Git, Docker, Kubernetes); experience with Ray, Airflow, or Dagster.
- Alignment Techniques: Practical experience implementing RLHF, RLAIF, DPO, or Constitutional AI pipelines at scale.
- Communication: Ability to distill complex technical trade-offs to cross-functional stakeholders.