Jayanth Varma

• Built a Microsoft Teams-integrated RAG chatbot using LangChain and Azure OpenAI, enabling CPA teams to query tax documents and client engagement data through natural conversation • Designed n8n automation workflows for intelligent document ingestion, classification, and context-aware retrieval and reducing manual processing overhead for tax preparers • Prototyped GPT-4 powered document summarization and auto-classification modules to route uploaded client files and surface key data points within HubSync's platform • Developed agentic trigger workflows to automatically flag missing documents, approaching deadlines, and anomalies across active client engagements • Researched and implemented LLM evaluation frameworks for hallucination mitigation and output quality assurance across tax-domain corpora with strict auditability requirements

Engineered real-time fraud detection systems using LSTM, Autoencoders, and Graph Neural Networks, significantly reducing false positive rates on live banking transaction data • Built end-to-end computer vision pipelines for remote KYC verification — integrating face liveness detection, document OCR (Tesseract, CRAFT), and facial recognition (FaceNet, ArcFace) for automated customer onboarding • Implemented 3D CNN-based anti-spoofing and deepfake detection to safeguard the integrity of selfie-based KYC submissions • Developed an early-stage GenAI pipeline using GANs for synthetic training data generation to address class imbalance in fraud detection workflows (2023) • Prototyped an LLM-powered internal QA assistant for compliance and audit queries using GPT APIs with domain-specific prompt engineering • Managed full MLOps lifecycle: model versioning, A/B testing, automated retraining pipelines, and production monitoring on GCP and Kubernetes

Major Subjects are Mathematics 1A,1B, 2A, 2B, Physics, Chemistry.