Overview
Architect large language models that can process and generate natural language, designing neural network parameters trained on large quantities of unlabeled text data. This is a Subject Matter Expert role that also leads model architecture decisions and mentors junior team members.
What you'll do
- Design and structure advanced language models capable of understanding and generating human-like text.
- Create neural network architectures that efficiently process vast amounts of unstructured text data.
- Conceptualize model parameters and ensure models are scalable and effective for diverse applications.
- Lead the development and refinement of language model architectures to meet evolving project requirements.
- Collaborate with various teams to align model capabilities with project goals.
- Engage with multiple teams, contribute on key decisions, and provide solutions to problems that apply across multiple teams.
- Mentor and guide junior team members to enhance their understanding and capabilities in language model design.
- Optimize model performance and scalability across different computational environments.
- Integrate language models into broader systems in collaboration with cross-functional teams.
What you'll need
- Proficiency in Large Language Models (LLMs).
- Minimum 15 years of experience in Large Language Models (LLMs).
- 15 years full time education.
- Experience with designing and tuning neural network parameters for natural language processing tasks.
- Strong understanding of model training techniques using large-scale unlabeled text datasets.
- Familiarity with state-of-the-art architectures and advancements in language model research.
- Ability to optimize model performance and scalability across different computational environments.
- Capability to collaborate effectively with cross-functional teams.
- Influencing and advisory skills.
Details
- Location: Pune office.
- Educational qualification: 15 years full time education.
Read the full description and apply on the company’s own careers page.