Logo of Huzzle

2024 Intern: Research Scientist – Conversation AI Model Development

image

IBM

Oct 11, 2023

Applications are closed

  • Internship
    Full-time
    Off-cycle Internship
  • Software Engineering
  • $104.4K - $143.5K
  • San Francisco

Requirements

  • Applicants should be PhD & MS students
  • Design, validation, and characterization of algorithms and/or systems
  • Machine learning engineering: creating training pipelines and evaluating models using toolkits such as PyTorch, TensorFlow, and scikit-learn
  • Preferred Technical and Professional Expertise:
  • Programming lanaguages: Python, Java, C/C++, JavaScript, R, etc.
  • Experience in training large-scale machine learning models
  • Experience analyzing large-scale data from a variety of sources
  • The candidate is expected to have 2+ years experience in NLP, machine learning, or computational linguistics and strong programming skills.
  • We prefer candidates with a strong publication record in conferences such as NeurIPS, AAAI, IJCAI, ICML, ICLR, ACL, EMNLP, and ICASSP.

Responsibilities

  • This is for a 2024 summer internship with the following start dates: May – August or June – September for quarter system schools.
  • IBM Research is looking for strong PhD-level interns to join our team in 2024 to work in the area of conversational AI model development. Our team pursues conversational AI from a number of different directions, and we have a range of possible internship projects, including the following options:
  • Explore novel approaches towards improving the performance of large language models in the context of enterprise use cases for conversational AI both for generation and retrieval. You will work with the Research team on aspects that may include synthetic training data generation, human in the loop curation of the data, model training, multilingual aspects and evaluation methods to assess the model performance on a variety of domains.
  • Multilingual Retrieval Augmented Generation, where you would design, build, and debug models that are able to answer questions in multiple languages by providing a answer that is consisted with a set of base documents containing the answers, but also detect when the question is outside the scope of the document collection. The research will cover building similarity models, performing search reranking, and generation in provided languages.
  • Explore the creation of large multimodal text/audio/speech models and their application to the creation of synthetic training data for speech-to-text models. The candidate for this project is expected to have research experience in generative modeling for speech, audio, or text, and ideally will have experience with multimodal language models.
  • Understand and explore the abilities of large language models for calling tools and enterprise APIs in a conversational setting. This research may include coming up with novel approaches for synthetic data generation, distillation approaches to smaller models, and exploring instruction fine tuning for improved performance on API/tools based tasks.

Technology
Industry
10,001+
Employees
1911
Founded Year

Mission & Purpose

At IBM, we do more than work. We create. We create as technologists, developers, and engineers. We create with our partners. We create with our competitors. If you're searching for ways to make the world work better through technology and infrastructure, software and consulting, then we want to work with you. We're here to help every creator turn their "what if" into what is. Let's create something that will change everything

Get notified when IBM posts a new role

Get Hired with Huzzle

Discover jobs with AI-powered precision. Autofill and track applications, create tailored resumes, and find the best opportunities across the web – all by simply chatting.

Already have an account?