Hiya! 👋

My name is Kabeer Akande, and I'm a senior machine learning engineer at OVO!

💛 I love productionising ML products and building scalable ML pipelines
🔭 Formerly at BigDEAL, S2DS, MindfulChef, and UOE

🚀 (Some) highlights

Built MLOPS platforms with Airflow and Vertex AI pipelines @OVO
Built Alexa-type chatbot (Conversational-BIM) with AWS @BigDEAL
Developed a personalised recommender system (Digital Waiter) @Mindful Chef
Innovations led to three granted US patents and over 50 peer-reviewed publications
Hold GCP ML Engineer and HashiCorp Terraform Associate (003) certifications
Endorsed by Royal Academy of Engineering as UK Promising Talent
My team won the most innovative prize at 2019 S2DS programme
Grant winner at COMSOC IEEE ICC 2018 US conference
Technical volunteer at codeyourfuture

🚢 Open source projects

MLOPS Platform

MLOPS_Platform is a robust machine learning platform designed to streamline ML workflows on Google Cloud Platform (GCP) resources. Following best practices in MLOps and DevOps, it offers a structured and efficient approach to managing data loading, preprocessing, model training, evaluation, and deployment. It features automated scaling, testing, integration, documentation, and deployment. If you want to find out more, I have published a development roadmap on my Medium page.

Pipeline Visualization

Below is a visualization of the MLOPS Platform pipeline components as executed on Vertex AI:

Terraforming Dataform

Terraforming Dataform establishes the fundamentals of a single repo, multi-environment Dataform with least-privilege access control and infrastructure as code setup. Datailed explanation of the workflow is published on my Medium page.

Architecture

Machine Learning Models as Product (MLMaP)

MLMaP is a paradigm shift in ML modelling. It demonstrates building ML algorithms as software products thus, enabling data scientists and ML engineers to package models similar to how software developers and engineers ship software applications. It is a step towards MLOps implementation.

Named Entity Recognition App (NER)

NER is a Named Entity Recognition (NER) application. The app is built using Hugging Face's Transformers and Gradio, allowing users to input text and receive identified named entities. It utilizes a fine-tuned BERT model (koakande/bert-finetuned-ner) for high-accuracy entity recognition. You can try it out on my Hugging space