Specialised AI Engineer

About Nscale

Nscale is taking on the hyperscalers by building a vertically integrated GenAI cloud platform. We own the data centres, software, and applications that power today's AI stack using sustainable technology solutions. We thrive on a culture of relentless innovation, ownership, and accountability, where every team member takes pride in their work and drives it with excellence and urgency. As a Nscaler, you’ll build trust through openness and transparency, where everyone is inspired to do their best work.

Collaboration is key, and we work together swiftly and respectfully, embracing adaptability and resilience in all we do.

Nscale is looking for Senior / Staff AI Engineers to join our core AI team and build the systems that power our GenAI cloud platform.

This role sits at the heart of our AI services platform, designing and optimising distributed systems that power large-scale training, post-training, evaluation, and low‑latency, high‑throughput inference under strict performance and efficiency constraints.

You may specialise deeply in areas such as inference optimisation, large‑scale training, post‑training (fine‑tuning, alignment), or evaluation systems
, or operate across multiple parts of the stack. In all cases, you’ll work on hard systems problems at scale
, where performance, efficiency, and developer experience are critical.

This is a hands‑on role for engineers who want to push the boundaries of how AI systems are built, optimised, and consumed by other AI engineers.

• Design, build, and optimise scalable AI platform systems spanning one or more of the following:
  • Drive inference performance and efficiency, including:
    • KV cache management
    • Continuous batching
    • Speculative decoding
    • Quantisation (INT8/4, FP8)
    • Sparsity, pruning, and model compression
  • Build and improve post‑training services, including:
    • Fine‑tuning (LoRA, QLoRA, adapters, full fine‑tuning)
    • Alignment (RLHF, DPO, reward modelling)
    • Agentic RL (tool calling, off‑policy training, parallel thinking, decoupled sampling and updating)
    • Dataset curation and data processing workflows
  • Develop evaluation and benchmarking systems to measure:
    • Model quality, safety, and regression
    • System performance (latency, throughput, cost)
    • Real‑world behaviour and feedback loops
• Develop and optimise distributed systems for GPU/accelerator workloads, focusing on scalability, reliability, and efficiency.
• Conduct performance analysis and bottleneck investigations across multiple components and stacks spanning training, post‑training, and inference.
• Collaborate with research, infrastructure, and product teams to build the right platform components based on customer demand and industry direction.
• Build developer‑facing APIs, SDKs, and tooling that enable other engineers to effectively use Nscale’s AI services.

• 5+ years of experience building production systems in machine learning, distributed systems, or high‑performance infrastructure.
• 4+ years of hands‑on experience in at least one core area, within large‑scale, production AI environments (e.g., AI labs, hyperscalers), such as:
  • Inference optimisation
  • Large‑scale training / pre‑training systems
  • Post‑training (fine‑tuning, alignment, distillation)
  • Evaluation and benchmarking frameworks
• Strong hands‑on expertise in at least one of the above areas, with working knowledge across others.
• Proven ability to design, optimise, and operate systems at scale, with a strong understanding of performance trade‑offs across latency, throughput, cost, and model quality.
• Deep understanding of transformer architectures, LLMs, and/or multimodal models, including their behaviour in production systems.
• Strong proficiency in Python and PyTorch, with a track record of building production‑grade ML systems.
• Experience with distributed compute and training paradigms (e.g., data/model parallelism, sharding, scheduling).
• Experience working close to the hardware/software boundary, such as:
  • GPU/accelerator optimisation (CUDA, ROCm, or similar)
  • Memory management and system‑level performance tuning
• Experience building or operating production inference or training systems at scale.
• Ability to design clean abstractions, APIs, and…