AI Platform Engineer, Training and Inference

Saviynt•San Francisco, CA

About The Position

Saviynt's AI-powered identity platform manages and governs human and non-human access to all of an organization's applications, data, and business processes. Customers trust Saviynt to safeguard their digital assets, drive operational efficiency, and reduce compliance costs. Built for the AI age, Saviynt is today helping organizations safely accelerate their deployment and usage of AI. Saviynt is recognized as the leader in identity security, with solutions that protect and empower the world's leading brands, Fortune 500 companies and government institutions. The AI Platform team is building the compute layer that trains, evaluates, and serves every AI model at Saviynt. We need an ML Platform Engineer to own distributed training on Ray + H100s, the multi-engine LLM inference mesh (vLLM, SGLang, NVIDIA Triton), and the full model promotion lifecycle — from shadow mode through canary rollout to GA. The AI Platform team's mission is to build a secure, scalable, product-agnostic AI foundation that enables Saviynt's identity products to deliver measurable AI-powered outcomes. Training & Inference is the engine — it turns data into deployed models that make Saviynt's products smarter.

Requirements

Experience in ML engineering with time in an ML platform or MLOps role
Production Ray depth: Ray Train, Serve, Core, and Data — debugged real production failures including NCCL timeouts, Plasma OOM, and Serve autoscaling lag
LLM serving engines: hands-on with vLLM, SGLang, or NVIDIA Triton — PagedAttention, prefix caching, and continuous batching tuned for latency/throughput targets
Distributed training: DDP, FSDP, NCCL collectives, gradient checkpointing, and mixed precision (BF16/FP8)
RL working knowledge: PPO, policy gradient, or RLHF — able to translate an algorithm into distributed compute primitives
Model lifecycle operations: MLflow registry, shadow/A/B/canary patterns, and auto-rollback on golden signal degradation
Vector databases: Pgvector or Qdrant — ANN index strategies, embedding upsert, and query latency tuning under inference load
Strong Python and PyTorch; Flyte or equivalent ML orchestrator
Bachelor's degree in Computer Science, Engineering, or a related field, or equivalent practical experience or equivalent military experience

Nice To Haves

Quantization (nice to have): INT8/INT4/FP8 post-training quantization (GPTQ, AWQ, or bitsandbytes)

Responsibilities

Own the Ray ecosystem end-to-end: manage KubeRay on GKE, tune Ray Core Task/Actor scheduling, operate the Plasma distributed object store, and configure Ray Data for GPU-direct streaming from GCS/S3
Operate distributed training with Ray Train: configure TorchTrainer + DDP/NCCL for multi-node H100 clusters, manage checkpoint lifecycle, implement spot-preemption recovery, and integrate warm-start fine-tuning for retrain pipelines
Build and operate the LLM inference mesh with Ray Serve: compose vLLM (PagedAttention), SGLang (RadixAttention), and NVIDIA Triton (TensorRT/ONNX) as a unified deployment graph with Plasma zero-copy memory sharing
Optimise inference performance: configure fractional GPU allocation, enable continuous batching, implement per-engine autoscaling based on request queue depth, and tune KV-cache block sizes
Design and operate the model routing layer: capability-based, version-based, and tenant-based routing with cost-aware fallback between self-hosted SLMs and cloud LLMs
Build RL training infrastructure: define Flyte workflows for RL pipelines (rollout, reward shaping, policy update, evaluation), integrate Ray RLlib or custom PPO/GRPO loops with Ray Train, and manage replay buffer persistence on GCS
Operate the full model promotion lifecycle: quality gate → integration tests → load tests (k6) → shadow mode → A/B gate → canary (10%→100%) with golden-signal auto-rollback
Operate the retrain pipeline: drift detection triggers, warm-start retraining, relative quality gates (V2 >= V1 − 2%), and automated Flyte DAG through to canary
Integrate RAG retrieval into the inference mesh: vector similarity search, context assembly, and prompt construction before LLM inference