Fraud, AML & Risk
Real-time fraud monitoring, anti-money-laundering and adverse-media models, built to be explainable and audit-ready for compliance teams.
Lagos, Nigeria · Open to opportunities
Hi, I'm Joseph Edet.
> Data Scientist_
Data Scientist & ML Engineer with 4 years' experience turning large, complex datasets into trading, commercial and fraud-prevention decisions — from a live high-frequency trading system to production fraud & AML detection.
01.
About me
I'm a Data Scientist & Machine Learning Engineer with 4 years of experience analysing large, complex datasets to inform trading and commercial decisions. I develop, deploy and maintain predictive models — time-series, regression and tree-based methods — including a live high-frequency trading system with a 70% signal hit-rate.
Today I build the machine learning behind Fraudspect, a fraud and AML compliance platform. I've also shipped AI/LLM and NLP systems, from automated feedback assistants to natural-language querying, and I automate end-to-end data workflows with Python, SQL and Apache Spark, deployed through MLflow and CI/CD.
I'm currently completing an MSc in Financial Engineering at WorldQuant University. I'm known for turning ambiguous, high-impact business problems into reliable, data-driven solutions alongside traders, product teams and commercial stakeholders.
View full resume4+years in data science & ML
70%signal hit-rate on a live HFT
system
12A/B tests & causal studies
led
1,500+data scientists & analysts
mentored
Trading & Quant
Signal generation, backtesting and live deployment with position sizing, drawdown limits and stop-loss controls. I also do time-series and stochastic modelling.
Commercial Analytics
Pricing optimisation, segmentation, lead scoring and propensity models, backed by A/B tests and causal inference that change real business numbers.
Applied AI & LLMs
RAG, LangChain and NLP systems, including automated feedback assistants, natural-language querying and fine-tuned intent classifiers.
02.
Tech stack
Predictive Modelling
- Time-series forecasting
- Regression
- Random Forest
- Gradient Boosting
- XGBoost / LightGBM
- Classification
- Risk modelling
- Propensity modelling
- Isolation Forest
- SHAP
Applied AI, GenAI & NLP
- LangChain
- RAG
- LLM decision support
- Prompt automation
- Hugging Face Transformers
- LoRA fine-tuning
- PyTorch
- TensorFlow
- Clustering
Trading & Quant
- Signal generation
- Backtesting
- Position sizing
- Risk controls
- Stochastic models
- Derivatives pricing
- Monte Carlo
- Portfolio optimisation
Data Engineering
- Python
- SQL
- Apache Spark / PySpark
- pandas
- NumPy
- scikit-learn
- ETL pipelines
- Data warehousing
- PostgreSQL
- Snowflake
- Oracle
MLOps & Cloud
- MLflow
- CI/CD
- FastAPI
- Model versioning
- Monitoring
- AWS
- Azure AI
- Git
Experimentation & BI
- A/B testing
- Causal inference
- Streamlit
- Power BI
- Matplotlib / Seaborn
- Jupyter
03.
Featured projects
FraudSpect
Financial Risk Resolver
Fraudspect — Fraud & AML Compliance Platform
A fraud and AML platform for African markets. It covers real-time transaction monitoring, KYC, and sanctions, PEP and adverse-media screening, with case management and regulatory reporting on one dashboard. I build, deploy and maintain the machine learning models that run its fraud, AML and adverse-media detection.
Neural-Network Signals for Systematic ETF Portfolios
MScFE capstone: LASSO, LightGBM and LSTM classifiers forecast 5-day direction for 11 US sector ETFs; probabilities become volatility-scaled long/flat signals targeting 10% vol, benchmarked after costs on an expanding walk-forward (2008–2026).
Irembo Voice AI — Intent Classification
A multilingual intent classifier for government-service access that handles noisy, code-switched voice transcripts across four languages. AfroXLMR is fine-tuned with LoRA, with back-translation augmentation, a "data clinic" that audits label quality, and a FastAPI inference service.
Anomaly Detection of Fraudulent Fintech Transactions
Detection system for fintech transaction logs using Isolation Forest and DBSCAN, with robust log parsing and SHAP-based explainability to surface suspicious user behaviour.
Detecting Credit Card Fraud in E-commerce
Comparative study of clustering, deep learning (CNN) and tree models (Random Forest, XGBoost) with PCA features and class-imbalance handling to find the best approach to e-commerce card fraud.
Calibrating Stochastic Volatility & Interest-Rate Models
Calibrates Heston to option data via Lewis (2001) and Carr-Madan (1999) Fourier pricing, extends to Bates (1996) jumps, then fits CIR to the Euribor term structure and simulates future rates.
Challenges in Financial Time-Series Models
Tackles non-stationarity (unit roots, cointegration, ECMs on Visa/Mastercard), Markov regime switching and multicollinearity (SPY/IVV/VOO) — derived, estimated, diagnosed and documented.
Predicting Stock Market Movements with Neural Networks
Replication of Sagaceta-Mejía et al.'s method for emerging markets using optimised technical indicators and neural networks — the foundation later extended in the ETF capstone.
Anomaly Detection on Offshore Platform Sensors
Flags operational anomalies in multivariate sensor data using Mahalanobis distance, PCA reconstruction error and Isolation Forest, with duration rules to separate incidents from a "new normal".
Crypto Price Movement Prediction
Classification model (Bitgrit competition) predicting whether a crypto asset's price rises or falls over a two-week window.
GenoSQL-AI
An open-source framework that combines PostgreSQL hybrid data models with AI-powered natural-language querying. It turns plain-English questions into structured data access for scalable, low-cost genomic analysis.
Price Surge in the Nigerian Economy (Omdena)
Led the data-preprocessing phase of an Omdena collaborative project visualising the recent price surge in Nigeria, shipped as an interactive Streamlit dashboard.
Visualising Apple's Carbon-Neutrality Progress
Maven Environmental Challenge entry: a data-journalism report on Apple's progress toward carbon neutrality by 2030.
04.
Experience & education
-
Aug 2026 — Present
AI & Machine Learning Engineer · Platview Technologies
- Build, deploy and maintain the production machine learning behind Fraudspect. The models cover real-time fraud detection and monitoring, anti-money-laundering (AML) detection, and adverse-media screening.
- Lead the AI and ML strategy for Platview's security products. I design explainable, audit-ready detection systems with feedback loops, so cases that analysts confirm or clear go back into retraining the models.
-
Jan 2024 — May 2026
Lead Data Scientist · PSP Analytics
- Designed ML models for pricing optimisation, customer segmentation and lead scoring, lifting conversion by 12% and shaping CVM and retention decisions.
- Engineered Python, SQL and PySpark pipelines across 6 business units, cutting data prep and training time by 40%.
- Led 35 A/B tests and causal-inference studies that drove an 8% increase in average order value.
- Built MLflow MLOps pipelines with CI/CD, versioning and monitoring, cutting deployment time by 65% and incidents by 45%.
- Shipped FastAPI and Streamlit apps that cut insight turnaround time by about 30%. Mentored 15+ junior data scientists and analysts, cutting onboarding time by 30%.
-
Jan 2026 — Jun 2026
ML Engineer (Quantitative Developer) · FloatEdge Capital
- Developed and deployed a research-driven high-frequency trading system on XAUUSD that achieved a 70% signal hit-rate in live trading.
- Built and backtested time-series and signal-generation models on 10 years of market data, then took the system live for real-time trading decisions.
- Implemented risk controls (position sizing, drawdown limits and stop-loss logic) and built low-latency data pipelines that cut inference and execution latency by 20%.
-
Mar 2025 — Jul 2025
AI / Machine Learning Engineer · The Factory Place
- Architected and deployed production AI systems for personalisation, segmentation and automated decision support.
- Built an NLP/LLM automated feedback and grading assistant that cut manual review workload by 40%.
- Developed a career-path recommendation engine using clustering and pattern recognition.
-
Apr 2023 — Jul 2024
Data & Product Analyst · PutHand
- Architected end-to-end data infrastructure (ingestion, normalised schemas and a warehouse layer) that powers 32 downstream reports.
- Built Isolation Forest fraud-detection models with 92% recall and 89% precision, preventing an estimated ₦560K (42%) in fraud losses.
- Automated Python and SQL pipelines that process high-volume daily transaction data.
// education
Jan 2025 — Present
MSc Financial Engineering (in view)
WorldQuant University
Quantitative risk modelling · Credit risk · Time series · Stochastic processes · Optimisation · Derivatives pricing · Monte Carlo · Portfolio optimisation
Nov 2015 — Jun 2021
BSc Pure & Industrial Chemistry
University of Nigeria, Nsukka
05.
Certifications & awards
Oct 2023
1st Runner-Up — DataFestAfrica 2023 Datathon
Built a fraud detection model and an end-to-end data pipeline (Snowflake → transform → Power BI) in a 3-person team, then presented the findings under a tight deadline.
Azure AI Fundamentals
Microsoft
Foundations of Financial Engineering
WorldQuant University
Fundamentals of Deep Learning
NVIDIA
Applied Data Science
WorldQuant University
ISO/IEC 42001 Implementer
PECB
IBM Data Analyst Specialization
IBM · Coursera
06. What's next?
Let's build something together
I'm open to data science, ML engineering and quant roles, as well as collaborations on fraud, risk and finance problems. My inbox is always open.
joebasshd@gmail.com