Описание:
Cast AI is an automation platform that operates cloud-native and AI infrastructure at scale. It uses autonomous decision-making in Kubernetes and cloud environments to continuously optimize production performance, reliability, and efficiency. Its Kimchi system automatically matches workloads to cost-efficient, high-performing LLM and serving configurations on customer infrastructure.
Задачи:
- Increase throughput through continuous batching, speculative decoding, chunked prefill, and kernel-level tuning across vLLM, SGLang, and TensorRT-LLM
- Reduce TTFT and TPOT latency by profiling bottlenecks and addressing compute, memory bandwidth, scheduling, or networking issues
- Improve KV cache utilization through paged attention, prefix caching, eviction policies, cache reuse, and quantized KV
- Quantize weights, activations, and KV using INT8, INT4, and FP8 while measuring quality on real workloads
- Reduce cold starts and memory footprint through faster initialization, smarter weight loading, and tighter memory accounting
- Scale inference across nodes using distributed topologies, network-aware placement, and checkpointing strategies
- Set the technical direction by deciding what to benchmark, adopt, and build, and align the team through writeups and reproducible experiments
Требования:
- 5+ Years building production ML systems
- A portfolio demonstrating depth in inference or training infrastructure, not just model training notebooks
- Strong Python skills for production services
- Hands-on experience with at least one of vLLM, SGLang, or TensorRT-LLM and an understanding of inference engine performance on different GPUs
- Fluency with quantization tradeoffs, including measuring quality regressions
- Experience with distributed systems, collective communication, sharding strategies, and failure modes in multi-GPU and multi-node setups
- A measurement-focused approach: instrument before optimizing and distinguish real improvements from benchmark artifacts
- Self-direction and comfort with a broad mandate
Условия:
- Competitive salary depending on experience level
- Equity options 10%
- Of work time for personal projects or self-improvement
- Learning budget for professional and personal development, including international conferences and courses
- Annual hackathon
- Team-building budget and company events
- Equipment budget
- Extra days off
- No visa sponsorship or work permit is provided
- A background check may be conducted at the final recruitment stage through Checkr