Intern – Speech Processing / ASR for Low-Resource Languages

Leidos

About The Position

The Intel Sector at Leidos is seeking a Technical Intern to support the Cyber & Information Sciences Division starting in Summer 2026. This is an exciting opportunity to contribute to the development and prototyping of innovative technologies that advance the state of the art in language and speech processing. This internship will support research and development efforts focused on audio enhancement, speech activity detection, and automatic speech recognition (ASR) for low-resource languages. The work will involve real-world audio collected in naturalistic environments, including recordings with variable quality, background noise, overlapping speakers, and inconsistent microphone placement. The intern will assist with developing methods and prototype pipelines to ingest, clean, segment, and analyze noisy audio to identify usable speech segments and candidate data for downstream ASR training and model bootstrapping. Areas of work may include audio enhancement, speech activity detection, speaker diarization, speaker identification, ASR, and data preparation techniques for low-resource language settings. The ideal candidate is curious, adaptable, and collaborative, with foundational experience in Python and an interest in speech, audio, or machine learning. This candidate will work closely with research scientists, data scientists, and engineers in support of an applied R&D project, and should be comfortable learning new tools, communicating progress, and contributing to technical problem-solving as part of a team.

Requirements

Current advanced undergraduate (rising junior or rising senior) or graduate student studying Computational Linguistics, Speech and Hearing Sciences, Acoustics, Electrical Engineering, Computer Science, Data Science, or a related technical field.
Coursework, academic project experience, lab experience, or other hands-on experience in one or more of the following: speech processing, audio signal processing, machine learning, natural language processing, acoustic phonetics or a related area.
Experience programming in Python through coursework, research, projects, or internships.
Strong problem-solving skills and the ability to communicate technical results clearly.
U.S. Citizenship required.

Nice To Haves

Coursework, project work, or research experience in one or more of the following: audio enhancement, speech activity detection, speaker diarization, speaker identification, ASR, or low-resource language technologies.
Familiarity with speech, audio, or machine learning tools and frameworks such as PyTorch, TensorFlow, torchaudio, Kaldi, ESPnet, SpeechBrain, Hugging Face, or similar.
Experience working with noisy, conversational, far-field, or otherwise real-world audio data.
Familiarity with data preparation, annotation workflows, corpus curation, or model bootstrapping approaches.
Familiarity with AWS or other cloud-based development environments.
Publications, class projects, research posters, GitHub projects, or other demonstrable technical work in a relevant area.

Responsibilities

Support Leidos engineers and researchers in defining tasks, workflows, and integration requirements.
Assist with designing and implementing a prototype pipeline to transform noisy audio into data usable by downstream speech models.
Help preprocess, enhance, segment, and analyze noisy audio for low-resource language applications.
Run experiments, evaluate results, and document findings.
Prepare technical content and communicate progress, challenges, and next steps in team meetings, design reviews, and technical exchanges.
Contribute to testing, prototyping, and iterative improvement of speech and audio processing workflows.