Closed listing
Senior LLM Research Engineer
- Toronto, ON, Canada
Full-time
This opening has closed.
The original listing is kept below for reference.
Job details
- Job type
- Full-time
- Closing date
- Posting language
- English
- Posted
Similar jobs hiring now
Search current openingsOriginal job description
Archived listing. The details below describe a past opening.
Senior LLM Research Engineer
Full-Time Permanent
Enterprise Client - Downtown Toronto
Hybrid/Remote
We're looking for a Senior LLM Research Engineer to help build and optimize advanced language models for complex, real-world applications. This role spans the full lifecycle of model development, from dataset creation and post-training through inference optimization, evaluation, and continuous improvement.
You'll work on challenging domains where high-quality training data and ground truth are often unavailable, requiring creative approaches to data generation, labeling, evaluation, and model alignment. Success in this role comes from combining strong research capabilities with practical engineering execution.
What You'll Do
- Lead post-training workflows, including SFT, RL-based training, distillation, and model alignment
- Design and implement robust evaluation frameworks, benchmarks, and regression testing suites
- Build high-quality training and evaluation datasets from large, unstructured corpora, including synthetic data generation
- Train and optimize models on distributed GPU infrastructure, troubleshooting performance and scalability challenges
- Improve inference efficiency through quantization, parallelization, decoding optimizations, and serving performance tuning
- Analyze results and translate experiments into data-driven model improvement decisions
What You'll Bring
- Strong Python expertise and experience with modern LLM frameworks such as PyTorch, Hugging Face Transformers, TRL, DeepSpeed, vLLM, or similar tools
- Hands-on experience fine-tuning and aligning large language models using supervised and reinforcement learning approaches
- Deep understanding of distributed training, model scaling, and GPU optimization
- A rigorous approach to evaluation, experimentation, and measurement
- Ability to bridge research and engineering, delivering production-ready solutions with strong software development practices
Preferred Experience
- Creating labeled datasets from unstructured data, including synthetic data generation at scale
- Designing LLM evaluation methodologies, rubrics, and LLM-as-judge systems
- Inference optimization, including quantization, speculative decoding, and performance tuning
- Building AI solutions for complex, domain-specific tasks where answers require nuance, reasoning, and judgment
More about this job
Responsibilities
Lead post-training workflows, including supervised fine-tuning, reinforcement learning, distillation, and model alignment, while building datasets and evaluation frameworks. Train and optimize models on distributed GPU infrastructure, improve inference performance, and use experimental results to guide model improvements.
Requirements
Requires strong Python expertise and hands-on experience fine-tuning and aligning large language models with supervised and reinforcement learning methods. Candidates should understand distributed training and GPU optimization, apply rigorous evaluation practices, and be able to deliver production-ready solutions using modern LLM frameworks.
Skills
- Python
- Large Language Models
- PyTorch
- Hugging Face Transformers
- Supervised Fine-Tuning
- Reinforcement Learning
- Model Alignment
- Distributed Training
- GPU Optimization
- Dataset Creation
- Synthetic Data Generation
- Model Evaluation
- Inference Optimization
- Quantization
- DeepSpeed
- vLLM
Visa sponsorship
Not detected in the job text
Categories
- Technology
- Software
- Engineering
- Science & Research
- Data & Analytics
Keywords
- Python
- Large Language Models
- PyTorch
- Hugging Face Transformers
- TRL
- DeepSpeed
- vLLM
- Supervised Fine-Tuning
- Reinforcement Learning
- Distillation
- Model Alignment
- Distributed Training
- GPU Infrastructure
- GPU Optimization
- Synthetic Data Generation
- Dataset Labeling
- Evaluation Frameworks
- Benchmarks
- Regression Testing
- Inference Optimization
- Quantization
- Speculative Decoding
- Parallelization
- LLM-as-Judge
- Model Scaling
- Data-Driven Experimentation
- Production Engineering