Mofetoluwa Adeyemi

Training and optimizing Mixture of Experts (MoE) and multimodal (MLLMs) models for internal tasks. Building AI Agents to collect and filter training and evaluation datasets. Curating and maintaining benchmark tools and datasets with Git CI/CD pipelines and inference frameworks such as vLLM and OpenAI-compatible servers. Developing custom reasoning LLMs using Reinforcement Learning, specifically with GRPO (Group Relative Policy Optimisation).

Machine learning research on text-based information retrieval, multilingual and cross-lingual NLP. Led the CIRAL project at FIRE 2023 where community evaluations were carried on retrieval and reranking models. Training language models for first-stage retrieval and multi-stage reranking. Implementing Large language models such as GPT and Zephyr as rerankers on low-resource languages.

Led the implementation of a content-based recommender system for HairX, an application to help manage hair health. A similarity-based recommender was developed to suggest hair products and regimens based on user hair type and needs.