AI Researcher — Inference Optimization in Brazil at Jobgether

This position is listed on behalf of a partner company, who manages all applications and next steps. Our partner is looking for a AI Researcher — Inference Optimization based in Brazil.

This role offers the opportunity to advance the performance of large-scale machine learning models through cutting-edge inference optimization research.
You will work at the intersection of AI research, model architecture, systems engineering, and hardware-aware optimization.
Your work will directly influence latency, throughput, memory efficiency, and the cost of running sophisticated AI workloads.
You will design and evaluate innovative optimization techniques and translate research findings into production-ready systems.
The role combines hands-on experimentation with close collaboration across research and engineering teams.
You will benchmark inference workloads across modern hardware accelerators and identify opportunities for measurable performance gains.
This is an impactful opportunity to help shape efficient, scalable AI infrastructure for real-world production environments.

• 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.