Aspiring Data Scientist | AI Researcher | Cybersecurity Enthusiast

Passionate about pushing the boundaries of anomaly detection and generative AI applications. Currently exploring novel approaches in transformer-enhanced variational autoencoders for fraud detection with research publication potential. Experienced in end-to-end ML pipeline development, from data engineering to production deployment.

Core Expertise: Deep Learning • Anomaly Detection • Generative Models • LLM Integration • Network Security • Financial AI

Research-Grade Anomaly Detection with Publication Potential

• Innovation: Novel comparative analysis of traditional VAE vs transformer-embedded VAE architectures for financial fraud detection using PaySim dataset and extending the methodology to the IEEE dataset.

• Methodology:

  • Trained VAE exclusively on non-fraudulent data to establish normal behavior patterns
  • Implemented transformer embeddings to capture complex sequential relationships
  • Extended analysis to IEEE fraud detection dataset with significant performance improvements

• Technical Achievement: Achieved substantial improvement in detection accuracy through transformer-enhanced latent representations

• Impact: Potential for academic publication due to novel approach and strong empirical results

• Skills: PyTorch, Transformer Architecture, Variational Autoencoders, Research Methodology, Statistical Analysis

🔗 GitHub Repository | Status: Exploring further and preparing for publication

Cybersecurity Data Engineering & Open Source Contribution

• Simulation Setup: Orchestrated controlled DDoS attacks using VMware (attacker) targeting Windows system (victim)

• Data Engineering:

  • Captured legitimate traffic (ping commands) and malicious traffic (hping3 DDoS attacks) using Wireshark
  • Developed automated PCAP-to-CSV conversion pipeline
  • Created labeled supervised learning dataset combining both traffic types

• Open Source Impact:

  • Released data preprocessing code on GitHub for community use
  • Published comprehensive dataset on Kaggle for research community

• Business Value: Enables rapid prototyping of network intrusion detection systems

• Skills: Network Security, Wireshark, PCAP Analysis, Data Engineering, Feature Engineering

🔗 GitHub Repository | 📊 Kaggle Dataset

Production-Ready LLM Application with Advanced Features

• Enterprise Features:

  • Multi-Statement Intelligence: Comprehensive analysis across Balance Sheet, P&L, and Cash Flow statements
  • Structured Output Parsing: Pydantic models ensure validated, consistent responses
  • Risk Assessment Engine: AI-generated risk levels with visual indicators (🟢🟡🔴)
  • Professional Visualizations: Dynamic charts tailored to each financial statement type

• Technical Excellence:

  • Robust Architecture: Modular design with comprehensive error handling and fallback mechanisms
  • Security Implementation: Environment-based API key management
  • Export Functionality: Downloadable comprehensive analysis reports
  • File Compatibility: Supports both CSV and Excel with intelligent parsing

• Business Impact: Transforms hours of manual financial analysis into automated, actionable insights

• Skills: LLM Integration, LangChain, Pydantic, Streamlit, Financial Analysis, Production Architecture

🔗 GitHub Repository | Status: Production-ready with full documentation

• RAG Implementation - Advanced retrieval-augmented generation techniques
• Gemini for Developers - Production LLM integration and optimization
• Prompt Engineering - Strategic prompt design for optimal AI performance

• Generative AI Certification - Smart Bridge & Google Partnership Program

Focus Areas:

• Variational Autoencoders (VAE) & Generative Adversarial Networks (GAN)
• Large Language Models (LLM) & Fine-Tuning
• Prompt Engineering & LangChain Framework
• RAG Implementation & Vector Embedding

Research & Publication:

• Finalizing research paper on VAE+transformer architecture using PaySim and IEEE fraud detection datasets
• Comprehensive experimentation comparing performance metrics across both datasets
• Targeting publication in top-tier ML conferences or specialized fraud detection journals

• Generative models for anomaly detection in financial systems
• LLM applications in automated risk assessment
• Intersection of cybersecurity, cloud and AI/ML

• Email
• GitHub
• LinkedIn
• Credly

Open to collaboration opportunities in research, industry projects, and cutting-edge AI applications.