WBG Pioneer - LSMS Survey Innovation Fellow

The Development Economics Survey Unit (DECSU), in the World Bank Group’s Development Economics Vice Presidency, supports the production and use of high-quality household survey data in low- and middle-income countries through the Living Standards Measurement Study (LSMS). LSMS works with national statistical offices and other partners to design, implement, and analyze multi-topic household surveys that inform policy and research on poverty, livelihoods, agriculture, labor, welfare, and related development outcomes. DECSU is developing an AI-focused work program to improve survey implementation and data quality through recent advances in large language models and generative AI. This work includes using interview audio to create analyzable text through recording, transcription, translation, and integration with survey metadata; strengthening survey quality control and error detection; and exploring how open-ended responses can support simpler, more respondent-centric questionnaire design. Duties and Responsibilities The Fellow will contribute to the design and development of an end-to-end pipeline to capture high-quality audio from field survey interviews, transcribe and translate audio across languages, including low-resource languages such as Nepali, Bengali, and Swahili, and integrate the resulting text with survey metadata for quality control and analysis. The assignment will provide practical exposure to survey methodology, data quality monitoring, field experiment implementation, and applied research using LLMs and generative AI. The scope of work will include reviewing and synthesizing the landscape of available automatic speech recognition (ASR) and translation technologies relevant to low-resource languages used in survey contexts; supporting the testing and evaluation of transcription and translation model outputs, including the design and application of accuracy benchmarks; contributing to data preparation, cleaning, and structuring tasks that enable the integration of audio-derived text with structured survey metadata; and assisting in the documentation of pipeline components, field protocols, and evaluation results to support the broader research and development effort. In consultation with the research team, the Fellow may also contribute to research outputs and develop an independent research question using project data.