Ayodeji Ajayi

Fine-tuned advanced LLMs and built a data-analysis integration feature, improving the model’s ability to process and analyze data by 50%. Sourced and enhanced datasets for an 18% benchmark accuracy lift, implemented RAG-style retrieval workflows, and applied RLHF with 20,000+ user preferences to improve user satisfaction by 25% and reduce real-world error rates by 40%.

Founded and led an AI startup generating research papers from prompts, building and deploying the core FastAPI service using OpenAI GPT models. Achieved 99.9% uptime and supported up to 500 concurrent LLM requests, while implementing RAG workflows, prompt evaluation, and LLM observability for continuous quality improvements.

Built ETL/ELT pipelines with PySpark, Databricks, and SQL transformations, enabling integrated analytics and downstream ML workflows. Designed cloud-native Azure/GCP architectures and GPU-accelerated Triton inference, reducing latency by up to 90% and increasing throughput by up to 4900%, while cutting post-deployment issues by 30%.

Implemented ML for credit card fraud detection, reaching 92% accuracy and reducing false positives by 17% using anomaly-based methods. Built computer vision pipelines and deployed LLM-based conversational/RAG systems, while improving data security and resilience (35% fewer potential vulnerabilities and 20% better data resilience).

Delivered AI solutions that increased user engagement by 35% and reduced customer support inquiries by 12%. Implemented DataOps with MLflow, DVC, and CI/CD (Azure DevOps/GitHub Actions), and used Terraform/Ansible to reduce downtime by 30%, achieve 100% uptime, and accelerate deployments by 21% with 25% fewer deployment errors.

Developed AI solutions achieving 85% accuracy using TensorFlow, while tuning distributed Spark jobs to reduce processing time and cloud compute costs. Built production monitoring for drift and data quality and deployed forecasting systems that reduced forecasting errors by 20%, plus Python-based chatbots and LLM-augmented ML orchestration for improved customer acquisition.

Set up scalable Azure infrastructure and streamlined CI/CD pipelines, increasing development speed by 20% and reducing time-to-market by 15%. Implemented end-to-end AI pipelines (ingestion, preprocessing, inference, and API integration) and built RAG-style NLP pipelines, creating three virtual assistants that handled 5,000+ interactions and saved an estimated 150 man-hours.

Led backend infrastructure enhancements using Docker, Ansible, and GCP, improving deployment reliability and scalability by 40%. Built CI/CD workflows for automated model training and releases, and developed Retrieval Augmented Generation systems to improve factual grounding and JSON-compliant outputs.

Completed an MSc in Data Science at the University of East London.

Earned a BSc in Computer Science from Landmark University.