Freelance AI & Machine Learning Adversarial Financial Fraud Detection

Job Title:
Freelance Researcher – AI & Machine Learning for Adversarial Financial Fraud Detection

Job Purpose:

· Lead end-to-end research, design, and implementation of advanced AI/ML-based financial fraud detection systems under adversarial data poisoning scenarios.

· Replicate, evaluate, and extend state-of-the-art Transformer-based fraud detection models using real-world financial transaction datasets.

· Design and simulate controlled data poisoning attacks (label flipping, feature manipulation, subpopulation poisoning, and backdoor attacks) to assess system vulnerabilities.

· Develop a robust Transformer-based detection model integrated with a Reinforcement Learning (RL) adaptive defence layer to counter evolving adversarial threats.

· Deliver a complete, reproducible Ph.D.-level research output including implementation, experimental evaluation, and a Scopus-ready dissertation and publications.

Required Qualification:

· Master’s or Ph.D. in Artificial Intelligence, Machine Learning, Computer Science, Data Science, Cybersecurity, or related disciplines.

· Strong academic or applied research background in fraud detection, adversarial machine learning, or cybersecurity analytics.

· Demonstrated experience with deep learning architectures (especially Transformers) and reinforcement learning systems.

Tools to be Familiar:

· Programming: Python
· ML/DL Frameworks: TensorFlow, PyTorch
· Data Handling: NumPy, Pandas, Scikit-learn, CSV-based large-scale datasets
· Visualization & Analysis: Matplotlib, Seaborn
· Experimentation: Jupyter Notebook, reproducible ML pipelines
Required Experience

· Minimum 2–4 years of experience in applied machine learning, fraud detection, or adversarial ML research.

· Prior experience working with financial transaction datasets or tabular ML problems.

· Hands-on experience in replicating published research models and conducting comparative evaluations.

Required Knowledge/Skills

· Deep understanding of supervised learning, deep learning, and transformer architectures for tabular data.

· Strong knowledge of data poisoning attacks, adversarial ML threats, and defence strategies.

· Practical expertise in reinforcement learning formulation, reward engineering, and adaptive decision systems.

· Ability to conduct statistically sound evaluations, robustness testing, and sensitivity analysis.

· Strong analytical thinking, independent research capability, and meticulous documentation skills.

· Excellent communication and collaboration skills for working with supervisors and reviewers.

Job Description

· Collect, preprocess, and engineer features from a large-scale real-world financial fraud dataset (IEEE-CIS Fraud Detection dataset).

· Replicate and validate state-of-the-art fraud detection models from recent academic literature.

· Design and implement controlled adversarial data poisoning scenarios applied strictly to training data.

· Develop a Transformer-based fraud detection model robust to adversarial manipulation.

· Integrate a Reinforcement Learning–based adaptive defence layer to dynamically respond to poisoning patterns and concept drift.

· Design an ensemble defence framework combining Transformer predictions, RL decisions, and SOTA model confidence scores.

· Conduct comprehensive experimental evaluation under clean and poisoned data conditions using standard fraud detection metrics.

· Prepare complete, reproducible deliverables including source code, datasets, experimental logs, and documentation.

Contact person:
Gray-95661 33822

Job Types: Part-time, Freelance, Volunteer
Contract length: 1 month

Pay: From ₹10,000.00 per month

Work Location: Remote