I build fraud-detection ML systems that are explainable to stakeholders and reliable in production.
From model design to deployment, I focus on measurable impact: faster pipelines, fewer false positives, and maintainable architecture.
Three examples of how I turn model work into measurable business outcomes.
Problem: Fraud signals were fragmented across separate tools and hard to operationalize.
Approach: Built deployable models for handwriting verification, signature forgery, and field extraction with shared evaluation standards.
Outcome: Shipped 10+ production models and reduced handoff friction between research and engineering.
Problem: OCR latency constrained fraud-review throughput and customer responsiveness.
Approach: Reworked pipeline bottlenecks, optimized model ordering, and tuned queue processing based on production traces.
Outcome: Delivered a 75% processing-time reduction with faster downstream fraud decisions.
Problem: Enterprise clients faced elevated false positives affecting operations and trust.
Approach: Led direct diagnostics with stakeholders, identified root-cause thresholds/features, and shipped targeted fixes quickly.
Outcome: Stabilized client performance and improved confidence in production fraud decisions.
I'm a Data Scientist with 4+ years of experience turning research ideas into production systems. My focus is document fraud detection across computer vision, deep learning, and platform engineering.
I care about work that ships and stays useful: reducing processing time by 75%, improving model quality in the field, and building systems teams can maintain as requirements evolve.
Deep Learning, CNNs, Siamese Networks, YOLO, Random Forest, PyTorch, TensorFlow
OpenCV, Image Processing, OCR, Document Analysis, Feature Extraction
Python, Java, System Architecture, Performance Optimization, Design Patterns
Handwriting Verification, Signature Detection, Check Analysis, Anomaly Detection
I own end-to-end ML delivery for check fraud detection, from research to production rollout.
Bridged R&D and engineering by integrating analytical improvements into the core Java fraud platform.
Experience machine learning concepts through hands-on visualization
Build and visualize a neural network architecture. Watch data flow through layers in real-time.
Create unique algorithmic art with mathematical patterns. Each generation is one-of-a-kind.
Tune the decision threshold and see precision, recall, review load, and expected business cost update in real time.
Why it matters: threshold tuning is where model quality becomes real-world fraud capture vs. false-positive pain.
Specialization: Intelligent Systems
Focus Areas: Deep Learning, Machine Learning, Software Engineering, and Data Analytics
Graduated with Highest Distinction
Strong foundation in competitive programming and algorithmic problem-solving
Dallas-Fort Worth Metroplex, Texas