AI Research Engineer

About Deloitte

Deloitte is the largest fully Canadian-owned professional services firm committed to making impactful contributions that foster thriving communities, businesses, and individuals alike. Our vision supports a prosperous society by empowering organizations with innovative solutions and responsible governance. Based in Toronto, we pride ourselves on building a better future with Canadian clients across multiple sectors.

Role Overview

We are seeking a committed AI Research Engineer specializing in generative AI systems, who will work at the crossroads of advanced AI research and production engineering. This permanent, hybrid role requires designing, assessing, and deploying generative AI models that are reliable and create significant impact. Responsibilities include fundamental research, model evaluation, and practical software development to progress next-generation intelligent applications.

Key Responsibilities

• Partner with cross-functional teams including product managers and engineers to develop AI-powered solutions meeting technical and business objectives.
• Explore, prototype, and build generative AI applications blending non-deterministic language models with deterministic engineering methods.
• Establish frameworks and benchmarks to measure model quality, dependability, and business outcomes.
• Regularly produce reports detailing model accuracy, drift, and operational performance.
• Troubleshoot and refine generative AI applications using prompt engineering, reinforcement learning, fine-tuning, and solid software engineering methods.
• Train and adapt large language models utilizing Hugging Face Transformers and reinforcement learning fine-tuning via Hugging Face TRL.
• Manage distributed and multi-GPU training workflows with tools like DeepSpeed, Accelerate, and FSDP, optimizing throughput and latency with frameworks such as vLLM.
• Help design scalable MLOps and DevOps pipelines for model deployment, monitoring, and ongoing training.
• Ensure all data handling adheres to privacy, security, and ethical AI standards.
• Keep abreast of latest developments in large language models, reinforcement learning, multimodal AI, and generative AI, applying insights to refine solutions.
• Document technical processes and contribute to publications, patents, or open-source projects as applicable.

Team Description

Deloitte AI and Data is a team of AI/ML specialists focused on delivering transformational AI solutions with a strong emphasis on responsible and transparent AI for Canadian organizations. The group aims to empower clients to become insight-driven and AI-first enterprises that produce measurable business value.

Ideal Candidate Profile

• Minimum 3 years of experience in machine learning engineering, data engineering, or applied research in academic or industry settings.
• Proficient in Python programming with experience using PyTorch, TensorFlow, or JAX frameworks.
• Hands-on use of Hugging Face Transformers for training, fine-tuning, or inference tasks.
• Experience with Hugging Face TRL for reinforcement learning based fine-tuning approaches like PPO, DPO, GRPO, or RLAIF.
• Operational knowledge of multi-GPU and distributed training using DeepSpeed, Accelerate, or FSDP.
• Skilled in running inference for large models with vLLM or comparable optimized platforms.
• Familiarity with experiment tracking and reproducibility tools such as Weights & Biases (W&B) and MLflow.
• Understanding of MLOps including continuous integration, training, monitoring, and lifecycle management.
• Experience working with generative AI frameworks including LangChain, AutoGen (A2A), or MCP.
• Commitment to writing clean, maintainable, production-quality code.
• Exposure to cloud AI environments (GCP, AWS, Azure), with relevant certifications valued.
• Strong foundation in reinforcement learning, natural language processing, and generative model architectures such as transformers, diffusion models, and retrieval-augmented generation.
• Research contributions through publications, patents, or open-source participation are a plus.

Preferred Technology Skills

• Expertise with Hugging Face Transformers library for model lifecycle management.
• Use of TRL for reinforcement learning fine-tuning.
• Applying DeepSpeed, Accelerate, or FSDP for large-scale distributed training.
• Utilizing vLLM for efficient large model inference and serving.
• Competence in experiment management through W&B or MLflow.
• Development of generative AI applications with LangChain, AutoGen, or MCP.
• Primary use of PyTorch deep learning framework.

Desirable Qualifications

• Experience with reinforcement learning from human or AI feedback (RLHF/RLAIF).
• Open-source AI framework contributions.
• Knowledge of model scaling laws, benchmarking, and evaluation metrics.
• Interest in advancing trustworthy, explainable, and safe AI approaches.

Total Compensation

The position offers a salary range from 72000 to 138000 Canadian dollars annually, with eligibility for a bonus program. Deloitte benchmark salaries competitively, rewarding individual contributions and strengths. Benefits include paid vacation, $4,000 yearly mental health support, a $1,300 flexible spending account, firm-wide closure days, dedicated learning days, and a hybrid work structure promoting well-being.

Diversity and Inclusion Commitment

Deloitte fosters an inclusive workplace welcoming candidates of all backgrounds, including persons with disabilities and Indigenous peoples. Our recruitment is grounded in global talent standards using a blend of human and AI screening focused on key competencies while respecting privacy. We respect Indigenous territories and acknowledge ongoing commitments to reconciliation and environmental stewardship.