Abhinay Krishna Bodi

     

I engineer high-throughput, low-latency systems and train frontier deep learning architectures, bridging scalable backends and async engines (Python, Rust, C++) with modern AI research (Vision-Language Models, Multi-Agent Systems, GNNs).

• Mastercard | Data Scientist

  • Trained self-supervised and regression models to distill merchant signals for ranking, and deployed prompt-tuned Qwen systems for real-time transaction anomaly detection.

• NYU Courant | Teaching Assistant

  • Natural Language Processing and Large Language Models.

• Brane Enterprises | Data Scientist & Software Engineer

  • Built enterprise document understanding and retrieval pipelines with Spark and SQL, while developing asynchronous event-driven conversational backends.

• New York University (NYU Courant) | M.S. in Data Science (GPA: 3.8 / 4.0)
• VNR VJIET | B.Tech in Computer Science & Engineering (GPA: 3.9 / 4.0)

• satavahana Rust Tokio HFT

  • High-frequency options trading engine in Tokio Async Rust.
  • Streams 1.5GB+/4min tick feeds using lock-free DashMap storage, CPU-pinned workers, and an 8-strategy signal pipeline with Half-Kelly risk allocation.

• cheque_forensics PyTorch Qwen3-VL NVIDIA DGX

  • Two-stage VLM document forgery detection pipeline.
  • Combines a 655.8M C-RADIOv4-H backbone with 30B Qwen3-VL verification on NVIDIA DGX Spark (13.2x training speedup, 0.89 precision).

• ArcPay-Agentic-Financial-System Python Multi-Agent USDC

  • Agent-driven financial execution system translating natural language to on-chain USDC payments and equities trading.
  • Guarded by a deterministic GuardianAgent liquidity and whitelist risk layer.

• H-ARC_challenge Python LLM Reasoning Program Synthesis

  • Tackling the ARC AGI benchmark through closed-loop hypothesis generation and neural code synthesis with 32B Qwen 2.5 Coder (11% accuracy vs. 0% direct prompting baseline).

• creditRisk_Prediction_GNNs PyTorch Geometric FA-GNN LSTM

  • Feature Attention Graph Neural Network (FA-GNN) with LSTM temporal modeling for dynamic borrower graph risk prediction (0.77067 AUC + SHAP explainability).

• adaptive-pairwise-preferences Python Bayesian ML Active Learning

  • Bayesian latent factor model for sequential pairwise active feedback on the Netflix Prize dataset (100M+ ratings), reducing required feedback queries by 40% via information gain maximization.