Logo of Huzzle

Internship

Graduate Intern, AI Technologies

Logo of Comcast

Comcast

•

1mo ago

🚀 Off-cycle Internship

Washington, DC +1

⌛ Closed
Applications are closed

Off-cycle Internship

Software Engineering•Washington, DC, Philadelphia

Description

  • Comcast AI Technologies has a graduate student intern position available during Summer 2024 (minimum of 12 weeks, May through September). We are an innovative research group within Comcast’s Technology & Product organization with offices in Washington DC and Philadelphia. We do groundbreaking research to support the development of novel Artificial Intelligence and Machine Learning based products for Comcast’s millions of consumers. We are seeking a talented researcher to join our team in tackling applied challenges within the realm of training large language models and enhancing conversational AI. You will be part of a team focused on the development, training, and testing of large-scale models,

Requirements

  • The ideal applicant should currently be enrolled in a university PhD program (strong MSc students are also welcome to apply), with 2+ years research experience in one of the relevant areas and be able to obtain a work authorization in the US. Applicants should also have good programming and software development skills, and be comfortable working in an interdisciplinary, team-oriented, applied research environment. The expected output of an internship is a proof-of-concept implementation leading to a scientific publication and/or a patent application.
  • Relevant Work Experience:
  • 0-2 Years

Education requirements

Currently Studying
Masters
PhD

Area of Responsibilities

Software Engineering

Responsibilities

  • Fine-tuning Conversational Models for various customer-related use cases.
  • Evaluate performance metrics like coherence, engagement, and satisfaction for refinement.
  • Design methodologies for high-quality data collection in Comcast's domain.
  • Explore synthetic data generation to address scarcity issues.
  • Investigate reinforcement learning with human feedback for model improvement.
  • Design reward mechanisms for contextually appropriate responses.
  • Implement active learning methods to select instances for human feedback and model training.
  • Investigate biases in conversational AI, particularly in customer interactions.
  • Implement rigorous procedures to identify and mitigate hallucination errors.

Details

Work type

Full time

Work mode

office

Location

Washington, DC, Philadelphia