PhD Student | Financial Markets & Machine Learning | Atlanta, GA

Two complementary research threads applying ML to financial market security:

Thread 1: Graph-Based Detection — Gamma-Sieve

• Heterophilic GNNs (CARE-GNN, TFE-GNN) detecting fragmented manipulation in options-equity markets
• Novel graph schema: 4 node types, 10 edge types for financial transaction networks
• Key finding: graph topology detects multi-account coordination (+5.6pp AUC over LSTM), but attack-type-dependent — ensemble required
• Domain-shift calibration: 42.5% FPR on real data → 2.4% after fragmentation calibration
• ESWA paper in preparation

Thread 2: LLM Market Understanding — gex-llm-patterns

• Obfuscation testing methodology for validating LLM understanding of dealer hedging constraints
• Paper #1: "Validating Large Language Model Understanding of Market Microstructure Through Obfuscation Testing"
  • ✅ Published: IEEE BigData 2025 — 2nd International Workshop on Large Language Models for Finance (Macau, December 2025)
  • 71.5% detection rate, 91.2% predictive accuracy (242 trading days, 726 evaluations)
• Paper #2: "Validating LLM Structural Reasoning: Detecting Persistent Market Regimes Through Temporal Obfuscation"
  • ✅ Accepted: AIAI 2026 (IFIP, Springer) — camera-ready May 2026
  • 📄 Under Review: Journal of Risk and Financial Management (JRFM, MDPI) — submitted March 2026
  • 81.2% regime detection 2024 vs 12.1% 2020 (69.1pp separation, φ = 0.672, p < 0.0001), 2,221 evaluations, 0% false positives on controls

• Gamma-Sieve — Heterophilic GNN evaluation for adversarial manipulation detection (PhD dissertation)
• gex-llm-patterns — LLM pattern detection in gamma exposure analysis (IEEE BigData 2025 published, AIAI 2026 accepted, JRFM under review)
• AutoTrader-AgentEdge — Multi-agent trading platform with GEX regime integration
• GexVisor — Financial visualization platform for GEX analysis and research tools

• ImproveBloodPressureMeasurements ⭐ 9 — ML-based blood pressure measurement accuracy improvement

• Transcript-ClusterViz — Conversation clustering and visualization using NLP
• nverma42/Chatbot — Collaborative chatbot development project

• NLPIntro — Natural language processing exploration
• Network-Compression-Analysis — Fast Wavelet Transform compression for neural networks
• BotNet Traffic — Botnet traffic detection via federated learning

Primary: Python (99% of work)

Also Proficient: C#, C++, TypeScript, JavaScript, Java

Domains: GNNs, LLMs, Algorithmic Trading, Market Microstructure, NLP, Healthcare ML, Signal Processing

• 🎓 PhD (In Progress) — Financial Markets & Machine Learning
• 🎓 M.Sc. — Computer Science
• 🎓 B.Sc. — CGDD & SWE
• ☁️ AWS Certified Practitioner

• AIAI 2026 (Accepted, camera-ready May 2026) — IFIP International Conference on Artificial Intelligence Applications and Innovations
  • "Validating LLM Structural Reasoning: Detecting Persistent Market Regimes Through Temporal Obfuscation"
• Journal of Risk and Financial Management (JRFM, MDPI) (Under Review, submitted March 2026)
  • "Validating LLM Structural Reasoning: Detecting Persistent Market Regimes Through Temporal Obfuscation"
• IEEE BigData 2025 — 2nd International Workshop on Large Language Models for Finance (December 2025, Macau, China)
  • "Validating Large Language Model Understanding of Market Microstructure Through Obfuscation Testing"
• ESWA (In Preparation) — "Gamma-Sieve: Heterophilic GNN Evaluation for Adversarial Market Manipulation Detection"
• PhD Symposium 2025 — Testing LLM Structural Reasoning in Complex Systems

📝 Blog: Post Essentials 📧 GitHub: @iAmGiG