Data Scientist & LLM Engineer based in Manchester, UK.
I build production‑ready machine learning systems, scalable NLP pipelines, and agentic workflows that solve real business problems across payments, energy, utilities, and logistics.

My work spans predictive modelling, LLM fine‑tuning, NLP automation, MLOps, and experiment design, with a strong focus on delivering measurable impact, reliability, and operational efficiency.

• LLM Fine‑Tuning & Optimization
  GPT‑40, LLaMA 3.1, T5, BERT — specialising in domain adaptation, prompt engineering, and evaluation of model consistency.

• Agentic Workflows & Automation
  Multi‑step pipelines for document parsing, onboarding automation, and customer‑service augmentation.

• Predictive Modelling & Experimentation
  Churn scoring, CLV modelling, forecasting, uplift modelling, A/B testing frameworks.

• NLP Pipelines at Scale
  Document extraction, topic modelling, semantic search, transformer‑based classification.

Spec‑Gap on GitHub
Evaluated LLMs using Spec‑Gap, a framework for measuring how closely models follow explicit task specifications.
Designed experiments assessing instruction‑adherence gaps across GPT‑40, LLaMA 3.1, and T5.
Improved prompt‑design strategies and consistency metrics for high‑stakes automation workflows.

Production‑grade churn models deployed via Databricks, MLflow, and CI/CD.
Advanced feature engineering, class imbalance handling, and governance‑aligned traceability.

Fine‑tuned LLaMA 3.1 70B & GPT‑40 using LangChain.
Automated ID&V checks, reduced resolution time, and improved compliance alignment.

Automated extraction from 20,000+ insurance documents.
Built robust parsing pipelines improving onboarding efficiency and data integrity.

Transformer‑based text classification improving resolution speed by 22.73%.
Sentiment analysis + topic modelling integrated into Power BI dashboards.

XGBoost pipelines predicting equipment failures with 87% precision.
PyTorch/TensorFlow models achieving 91.4% accuracy in oil‑production forecasting.

Built a semantic search system using LDA, LSI, BERT, and Sentence Transformers.
Interactive web app for exploring topic‑driven recommendations.

Languages: Python, SQL
ML: scikit‑learn, XGBoost, LightGBM, PyTorch, TensorFlow
LLMs: GPT‑40, LLaMA 3.1, T5, BERT, DistilBERT
NLP: spaCy, BERTopic, Hugging Face, LangChain, LlamaIndex
Cloud: Azure Databricks, AWS, GCP
MLOps: MLflow, CI/CD, model monitoring
Analytics: A/B testing, funnel analysis, retention cohorts
Dashboards: Power BI, Looker, Tableau

MSc Data Science & Artificial Intelligence (Distinction)
University of Liverpool

BEng (Hons)
Covenant University

• Email: ifeoluwajayeolaa@gmail.com
• LinkedIn: (https://www.linkedin.com/in/ifeoluwa-jayeola/)