AI Research Engineer (Model Compression & Quantization) - 100% Remote Worldwide
Tether Operations Limited5 months ago
Remote, WorldwideSenior
Responsibilities
- Design and deploy state-of-the-art model-serving architectures for high throughput, low latency, efficient memory usage, and operation across resource-constrained and edge environments.
- Establish performance targets and monitor latency, throughput, memory consumption, error rates, and device-specific inference metrics.
- Build and run controlled inference tests in simulated and production environments, documenting results against established benchmarks.
- Prepare test datasets and simulation scenarios for evaluating model performance, latency, and memory utilization on low-resource devices.
- Analyze computational and memory bottlenecks, including batch-processing inefficiencies, network delays, and excessive memory usage.
- Integrate optimized serving and inference frameworks into production pipelines for edge and on-device applications.
- Design and optimize distributed inference engines using tensor parallelism, pipeline parallelism, and expert parallelism for GPU clusters.
- Iterate on inference optimization strategies and define success metrics for performance, reliability, scalability, and memory efficiency.
Requirements
- Bachelor’s degree in Computer Science or a related field; a PhD in NLP, Machine Learning, or a related field is preferred.
- Solid track record in AI research and development, ideally including strong publications at A* conferences.
- Knowledge of Metal Shading Language and ability to write custom compute shaders from scratch.
- Proven experience with low-level kernel optimization and inference optimization on mobile devices.
- Demonstrated improvements in inference latency, throughput, and memory footprint for domain-specific applications on resource-constrained or edge devices.
- Deep understanding of model-serving architectures, inference optimization techniques, and efficient memory management.
- Strong expertise writing GPU kernels for mobile devices and understanding model-serving frameworks and engines.
- Practical experience developing and deploying end-to-end inference pipelines on resource-constrained devices.
- Ability to apply empirical research to latency optimization, computational bottlenecks, and memory constraints.
- Proficiency designing evaluation frameworks and iterating on optimization strategies.
- Understanding of diffusion models, vision transformers, pruning, quantization, FlashAttention, KV cache, and speculative decoding such as Eagle.
Benefits
- 100% remote work worldwide with a globally distributed team.
Categories
AI ResearchML Engineering
About Tether Operations Limited
Tether Operations Limited develops and issues reserve-backed digital tokens used by exchanges, wallets, payment processors, and merchants, most notably the USDT stablecoin available across multiple blockchains. Founded in 2014 and privately held, it earns revenue from token issuance, reserve management, and related services, and has expanded into tokenization, peer-to-peer communications (Keet), and Bitcoin mining. USDT is widely used in crypto trading and cross-border transfers, including in emerging markets.