B2B AI Application:

• Multi-media text/image/video understanding

LLM inference optimization and responsible AI:

• Optimized LLM inference with sglang and vLLM, achieving 50% latency reduction
• Built multilingual PII and policy violation classifiers (Sparse RoBERTa, Qwen3-0.6B), improving precision by 60%
• Developed guardrail models for Alexa’s core LLM ensuring policy-compliant responses across languages

Clinical NLP research:

• Built clinical diagnosis reasoning via knowledge graph networks with LLM reasoning and response re-ranking
• Designed LLM pretraining pipeline over large-scale healthcare knowledge bases with distributed training
• Contributed to LogosKG: hardware-optimized scalable knowledge graph retrieval system
• Coauthored paper on clinical text summarization with temporal reasoning (EMNLP 2025)

AI-powered newsletter generation:

• Built event clustering/extraction on web-scraped data with RAG for news summarization
• Engineered LLM-powered summarization and rewriting pipeline for production content workflows

HPC, distributed systems & AI infrastructure research:

• Designed decentralized ML system for NASA spacecraft networks using blockchain-based coordination
• Built distributed ML on embedded devices (Raspberry Pi, Nvidia Jetson) using Torch-TensorRT
• Developed GPT-3.5 agentic coding workflow for autonomous robot control code generation
• Published multiple papers in IEEE and top-tier venues on distributed systems and cybersecurity

HPC storage systems:

• Optimized HPC distributed storage I/O throughput prediction using LSTM/RNN models on Lustre parallel filesystem across 32-node clusters

Networking security & smart grid systems:

• Designed HIL testing infrastructure for large-scale renewable energy systems
• Implemented attack detection and resilient communication middleware for smart grids using ML and SDN
• Developed blockchain-powered decentralized computing system (2 US patents granted)