M.S. in Data Science at NYU, working on LLM inference serving and GPU / cluster infrastructure. Contributor to the vLLM ecosystem.
- sglang-inference-optimization: PD disaggregation lifted SLO attainment 52.9% → 99.2%
- research-agent-framework: parallel sub-agent DAGs cut latency 57.2%
- cold-chain-monitor: Kafka → Postgres/Redis pipeline, breach detection p99 < 1 s
vLLM-Omni (4) · AIBrix (3) · Semantic Router (2)
- ReKV: Overlapping Quantized KV Transfer and Suffix Recomputation: ICASSP 2027 (under review)
- Motion Recency Maps for Zero-Shot Traffic Accident Localisation: CVPR 2026 AUTOPILOT Workshop
| Languages | Python Go Java C++ JavaScript TypeScript CUDA SQL |
| LLM inference | vLLM SGLang AIBrix Semantic Router Mooncake PD Disaggregation KV Cache Quantization Speculative Decoding FP8 |
| GPU & cluster | CUDA Kernels Triton NCCL InfiniBand RDMA Nsight Systems Nsight Compute Kubernetes Volcano Slurm |
| LLM & agents | LangGraph MCP Function Calling Multi-Agent Orchestration RAG Embeddings Reranking LLM Evals Guardrails |
| ML | PyTorch MMDiT DINOv2 LoRA Hugging Face |
| Backend & data | REST APIs Spring Boot FastAPI Node.js React PostgreSQL MySQL MongoDB Redis Kafka |
| Cloud & DevOps | AWS (EC2, EKS, Lambda, API Gateway, S3, IAM) Docker Terraform GitHub Actions CI/CD Linux |

