AI Researcher — Inference Optimization

Accountabilities:

• Research and develop advanced techniques to improve inference performance for large neural networks and machine learning models.
• Optimize key performance dimensions including latency, throughput, memory efficiency, and cost per inference.
• Design and evaluate model-level optimization techniques such as quantization, pruning, KV-cache optimization, and architecture-aware simplification.
• Implement systems-level optimizations including dynamic batching, kernel fusion, multi-GPU inference, and prefill versus decode optimization.
• Benchmark and profile inference workloads across different hardware accelerators to identify performance bottlenecks and optimization opportunities.
• Collaborate closely with engineering teams to integrate optimized inference techniques into scalable production pipelines.
• Translate research findings and experimental results into reliable, production-ready improvements.
• Establish clear benchmarks, document findings, and communicate results to inform technical and product decisions.
• Explore emerging approaches such as long-context inference, speculative decoding, KV-cache compression and paging, efficient decoding strategies, and hardware-aware inference design.

Requirements:

• Strong background in machine learning, deep learning, AI systems, or a closely related technical discipline.
• Hands-on experience optimizing inference workloads for large-scale machine learning or neural network models.
• Strong proficiency in Python and experience with modern machine learning frameworks such as PyTorch.
• Practical experience with inference and model-serving technologies such as Triton, TensorRT, vLLM, or ONNX Runtime.
• Ability to design rigorous experiments, interpret performance results, and communicate technical findings clearly.
• Experience deploying production inference systems at scale is highly desirable.
• Familiarity with distributed inference and multi-GPU architectures is a plus.
• Experience contributing to open-source machine learning or inference frameworks is advantageous.
• Peer-reviewed research publications in machine learning, systems, or related fields are a strong plus.
• Experience working close to hardware through technologies such as CUDA, ROCm, or performance profiling tools is beneficial.
• Strong analytical and problem-solving skills, with the ability to translate research concepts into practical engineering improvements.
• Familiarity with advanced inference topics such as long-context optimization, speculative decoding, KV-cache compression, efficient decoding, or hardware-aware model design is advantageous.

Benefits:

• Full-time opportunity within a research-focused AI environment.
• Fully remote work from India.
• Opportunity to work on large-scale machine learning models and high-performance inference systems.
• Exposure to advanced model optimization, systems engineering, and hardware-aware AI techniques.
• Opportunity to contribute to production systems where research can generate measurable improvements in latency, throughput, and cost efficiency.
• Hands-on experience with modern inference technologies and hardware acceleration.
• Opportunity to explore emerging research areas including speculative decoding, long-context inference, KV-cache optimization, and efficient decoding strategies.
• Collaboration with research and engineering teams working on challenging real-world AI performance problems.
• Opportunity to contribute to open-source machine learning or inference technologies where applicable.
• Direct impact on the reliability, scalability, and efficiency of production AI systems.