Deskripsi Pekerjaan
Nexus Intelligence Labs is at the forefront of building foundational models for enterprise automation. We are seeking a Senior Machine Learning Engineer to architect, train, and deploy large-scale models that power our next-gen AI platform. You will work on the bleeding edge of LLMs, retrieval-augmented generation (RAG), and multimodal systems, directly impacting how Fortune 500 companies leverage AI.
This is a high-autonomy role for a builder who thrives in a fast-paced, research-heavy environment. You will own the full lifecycle from data curation and distributed training to low-latency inference optimization.
Why Join Us?
- Compute Budget: Unrestricted access to H100/A100 clusters.
- Open Research: Publish at top-tier conferences (NeurIPS, ICML, ICLR).
- Equity: Pre-IPO stock options with quarterly liquidity events.
- Culture: Zero bureaucracy, async-first, deep-work focused.
Tanggung Jawab
- Design and implement scalable ML pipelines for pre-training, fine-tuning, and alignment of LLMs (7B–70B+ parameters).
- Optimize distributed training workflows using PyTorch FSDP, DeepSpeed, or JAX/Flax across multi-node GPU clusters.
- Develop novel RAG architectures and agentic workflows for complex reasoning tasks.
- Build robust evaluation frameworks (automated & human-in-the-loop) for model quality, safety, and hallucination detection.
- Collaborate with Product & Data Engineering to define data flywheels: curation, annotation, and active learning loops.
- Deploy models to production with sub-100ms latency using TensorRT-LLM, vLLM, or custom CUDA kernels.
- Mentor junior engineers and establish best practices for MLOps, experiment tracking, and reproducibility.
- Stay current with literature; reproduce and adapt SOTA techniques (e.g., MoE, Speculative Decoding, DPO/RLHF).
Kualifikasi
- M.S. or Ph.D. in Computer Science, Machine Learning, or equivalent practical experience (5+ years industry ML).
- Expert-level Python & PyTorch/JAX; strong C++/CUDA proficiency for kernel optimization.
- Proven track record training/fine-tuning LLMs at scale (multi-node, multi-GPU).
- Deep understanding of Transformer architecture, attention mechanisms, and scaling laws.
- Hands-on experience with MLOps stack: Kubernetes, Docker, MLflow/Weights & Biases, Kubeflow.
- Strong software engineering fundamentals (CI/CD, testing, design patterns, version control).
- Publications at top ML venues or significant open-source contributions (e.g., Hugging Face, vLLM, LangChain).
- Excellent communication skills; ability to distill complex research for cross-functional stakeholders.