AI/ML Data Scientist specialising in NLP, predictive modelling, and automated analytics – with growing exposure to electronic fixed income trading and enterprise-grade data workflows.

I care about one thing: turning messy real-world data into models and tools that change decisions, not just slide decks.

• Predictive modelling & retention / churn

  • Supervised models (XGBoost, Neural Networks, logistic regression) for student and client retention, early-warning risk flags, and intervention design.

• Time-series, anomaly detection & operations

  • Isolation Forest, One-Class SVM, and rule-based QC to detect failures in sensor data and live systems, aligned to operational-risk thresholds.

• NLP & applied AI

  • LLMs, RAG, FinBERT, topic modelling and summarisation for finance and analytics use cases (research, monitoring, internal tooling).

• Markets & trading

  • Exposure to Nomura’s eFI Quant Rates desk: probability-of-fill models, ensemble architectures (logistic regression + trees/NN), and “risk radar” concepts for intraday risk buckets.

Languages & Data

• Python (pandas, NumPy, scikit-learn, XGBoost)
• SQL
• Jupyter / VS Code

ML & DL

• XGBoost, Gradient Boosting
• Neural Networks (TensorFlow / PyTorch basics, Keras)
• Clustering (K-Means, Hierarchical)
• Anomaly detection (Isolation Forest, One-Class SVM)
• Bayesian thinking for uncertainty & risk

AI & NLP

• LLM fine-tuning, LangChain / LangGraph
• RAG pipelines, NER, topic modelling
• HuggingFace Transformers, FinBERT, summarisation, agentic workflows

Cloud, BI & Dev

• AWS (S3, Athena)
• Power BI (Microsoft Certified)
• Docker, Git, GitHub

• Student / customer retention models
  End-to-end workflows: feature engineering, model comparison (baseline vs XGBoost vs NN), AUC/recall trade-offs, and tiered intervention strategies. ]

• Anomaly detection in sensor and operational data
  Reproducible notebooks that combine models + QC frameworks, aimed at reducing false positives and quantifying maintenance or risk savings.

• Customer segmentation & clustering
  EDA + clustering to define actionable segments, with a strong focus on interpretability and impact on downstream marketing or product decisions.

• Trading & execution prototypes (WIP)
  Notebooks exploring execution modelling, ensemble architectures for probability-of-fill, and risk dashboards inspired by work on an electronic rates desk.

Each repo aims to show the full chain: from business question → data prep → modelling → evaluation → recommendations.

• Studying on the University of Cambridge PACE Data Science & AI programme (Level 7), sponsored by the Bank of England.
• Building a more opinionated portfolio around:
  • retention modelling,
  • anomaly detection,
  • trading and execution analytics,
  • and practical NLP for finance.

If you want to talk about applied ML in trading, risk, or operations, feel free to reach out (victoreigbefoh@outlook.com) or open an issue on any repo.