Lead Agentic Ai Researcher

Human Resources•Toronto, ON

2d•Hybrid

About The Position

Toronto’s AI Lab is seeking a Lead Agentic AI Researcher to drive the development and application of agentic AI systems across LG Electronics’ global product portfolio. This role will lead research efforts focused on enabling self-learning, autonomous, and personalized AI agents that power next-generation LG products, including smart home platforms, connected appliances, and future intelligent devices. As a technical leader, you will define research direction and architecture for agentic AI capabilities such as lifelong learning, hierarchical memory, autonomous planning, and personalized inference. You will work at the intersection of advanced research and real-world deployment, ensuring that novel agentic AI concepts are designed with scalability, reliability, and product impact in mind. In close collaboration with engineers, product teams, and external research partners, you will translate research breakthroughs into production-grade prototypes and scalable systems. Your work will help establish agentic AI as a core intelligence layer across LG’s product ecosystem, ultimately shaping how millions of users interact with intelligent devices worldwide.

Requirements

PhD in Machine Learning, Artificial Intelligence, Computer Science, or a related field with 3+ years of post‑graduate research experience, OR Master’s degree in ML/AI or related field with 6+ years of relevant post‑graduate experience
Proven research experience in areas such as agentic AI systems, reinforcement learning, continual learning, hierarchical memory, multi‑agent systems, or large language models
Deep expertise in agentic AI architectures, including autonomous planning, skill libraries, lifelong learning, and memory‑augmented systems
Strong hands‑on experience using PyTorch and/or JAX for large‑scale model training
Experience with distributed training frameworks (e.g., FSDP, Megatron, or similar)
Hands‑on experience designing RL reward functions for long‑horizon tasks, agent evaluation, and memory management
Strong ability to establish research baselines, evaluation protocols, and metrics for long‑context reasoning, planning, and skill retention
Experience tuning and optimizing large multimodal or foundation models and building robust evaluation pipelines
Strong publication record in top-tier conferences or journals (e.g., NeurIPS, ICML, ICLR, ACL) related to agentic AI, continual learning, or personalized AI systems
Demonstrated ability to define and lead focused research roadmaps under resource or capacity constraints
Experience leading or significantly contributing to academic collaborations, including co‑authoring papers and managing university partnerships
Track record of successfully transitioning research concepts into production-grade prototypes or systems
Passion for mentoring and developing junior researchers and fostering a strong research culture
Strategic thinker with the ability to balance long‑term research vision and near‑term practical impact
Highly creative problem solver with a demonstrated ability to generate novel, real‑world solutions
Comfortable working in a fast‑paced, technically complex, and collaborative environment
Strong communication skills, with the ability to clearly explain complex research ideas to both technical and non‑technical stakeholders

Responsibilities

Lead research on continual learning agents with explicit skill libraries/tools and curriculum/goal discovery.
Thrive in a fast-paced, technically challenging environment, where you can contribute your innovative ideas and solutions.
Design self-learning agent systems for lifelong skill acquisition/evolution; establish novel reward functions for agentic task completion quality and long-horizon consistency.
Develop personalized inference control via activation steering or persona vectors; research dynamic persona alignment ensuring user compatibility and enterprise safety.
Architect unified Agent–User–Memory frameworks integrating memory agents, user modeling (probabilistic embeddings, belief states), and orchestration layers for proactive personalization.
Lead university collaborations, co-author papers targeting A conferences.
Translate research into production-grade prototypes.
Establish research baselines and evaluation protocols for long-context reasoning, test-time planning, memory utility, and skill retention; harden against overfitting and distribution shift.
Enjoy tuning and optimizing large multimodal models and have experience building evaluations to measure their performance.