-Designed the data moat strategy for the company’s linguistic dataset. competitiveness and secured the first data partnership with access to five of South Africa indigenous languages. - Leading the expansion and execution of data strategy into East and West Africa. - Providing useful pipeline data for model finetuning, evaluation and training. - Expanding the data operations of the company with data achieving, curation, licensing and governance.

- Led the research on the curation of corpora and LLM model evaluation on Reading Comprehension for Yoruba, Igbo and Hausa. - Annotated over 3,000 text dataset of Nigerian Pidgin on a news domain Topic Classification task. - Worked as an annotator on a Google collaboration project on Question Answering. - Coordinated a team of 5 translators working on the translation of about 200 scientific research papers from English to Yorùbá language. - Led a team of 3 annotators who worked on a multi-modal translation and computer vision project. - Co-ordinated 6 annotators on a combined 9,500 text datasets who worked on Named Entity Recognition Annotation and Parts of Speech Tagging project. - Led a team of translators who translated 8,000 sentences from English language to Nigerian Pidgin English. - Performed human model evaluation of a model trained majorly on dataset from the news domain.

Led 2 voice artists working on recording text datasets as audio for text-to-speech. - Managed 80 volunteers on recording text datasets as audio for automatic speech recognition. - Curated, generated, coordinated and cleaned text datasets which are to be recorded and used as aligned text for the ASR and TTS project. - Coordinated the localization of web and android contents into Yoruba for Mozilla Common Voice using Pontoon.

Led the development and formulation of rule-based style guides to aid numerical model development in Yorùbá language. - Supplied the written and spoken form of contents needed for the project to develop text-to-speech tools. - Supervised the validation of speech and text corpus in grammatical correctness, dialectal difference, audio quality etc - Evaluating cardinal and ordinal numeral formats in relation to contextual use in the project. - Provided 50 contextual sample contents as training data for model evaluation. - Created rule based grammar for capitalised words on a speech recognition task.

- Worked on a Machine Translation task for Google with an aim of improving the Yorùbá language machine translation quality on Google Translate. - Identifying sentences from the source during translation which hinders target language quality of translation. - Using Google’s Transform 3 MT tool to optimise translation workflow.

- Contributed to the development of ontology and style guide which guides annotation of speech phonetically and documentation of annotation processes. - Lead a team of 5 annotators who annotated about 2000 dataset of speech corpus. - Tested a conversation design tool and gave recommendations on how to improve the tool. - Worked on evaluating a speech annotation tool and recommended some linguistic design features which made the annotation process swift by about 55% optimization.

- Documented terminology processes of over 10 partners for translation works. - Used automated and manual techniques to extract, clean and create terminology. - Worked mostly with Indo-European, African and American languages.

- Led 3 annotators on sentiment classification of 31,000 sentences dataset on a research grant for Yorùbá language. - Co-ordinated 3 annotators sentiment tagging of 27,000 sentences dataset in the Yorùbá language in the quest of ensuring low resource languages have a better representation in technology. - Instrumental in the development of style guides of unique features in African languages in relation to sentiment classification and tagging.

- Worked with the director of the project remotely and supplied about 700 names into the dashboard. - Edited and added meanings, morphology and geolocation to names in the dashboard. - Used and tested the text-to-speech feature aiding pronunciation of names in the dashboard.