- π CS student at Rutgers University, building full-stack & AI-powered apps
- π Currently working on Sous (a bilingual vLLM / SGLang deployment-command generator) and Neutoken BM Platform (an LLM benchmark console)
- π± Currently learning LLM inference & serving and system design
- π€ Robotics background with FTC Team 9121 Falcon Force (robot code, team website)
- π¬ Ask me about vLLM / SGLang deployment, Next.js + Flask full-stack, and ESP32 embedded projects
- π« How to reach me: Via GitHub Issues
Sous
 Bilingual (δΈ/θ±) deployment-command generator that merges the official vLLM recipes and SGLang cookbook generators into one app: 164 models, 160 vLLM recipes, 23 SGLang deploy panels, 2 mamba-ratio calculators, and 160 guide docs translated to Chinese. Does not rewrite any engine code. |
Neutoken BM Platform
 LLM benchmark console and dashboard: accuracy benchmarks (AIME2025, HLE, SWE-Bench) vs official baselines, TTFT P50/P90 vs context-length charts, light/dark/system themes, password-gated access, server-side masked API keys, and live SSE log streaming. |
Magic Glove
 Wearable BLE air mouse + wireless keyboard on a Seeed XIAO ESP32-S3 with an MPU-6050 6-axis gyroscope, tactile buttons and buzzer feedback. Turns hand gestures into HID input. |
AI Interrogation Room
 AI-driven detective mystery game: interrogate three suspects, each with independent dialogue history and a psychological-pressure system, in a dark glassmorphism UI. |
Falcon Force 9121 Website
Static marketing site for Falcon Force 9121, a FIRST Tech Challenge robotics team, live at falconforce.tech. |
Into-The-Deep
 Robot code for FTC Team 9121 Falcon Force, 2024-2025 "Into the Deep" season. |
