Wenhui Tu

Experience

1. Ai Engineer

   BlackRock

   Jan 2025 — Present · New York, NY, US

   Developed a FastAPI‑based microservice that generates hourly financial reports via a LangChain pipeline; provides a REST API for on‑demand retrieval with <200 ms p95 response time. • Designed a Retrieval‑Augmented Generation (RAG) pipeline to ingest and vector‑store real‑time news, FX rates, and market data; enabled secure offline querying through a private knowledge base chatbot with local embeddings. • Fine‑tuned large language models (DeepSeek‑R1‑Qwen‑14B & Llama‑2‑7B) using parameter‑efficient LoRA adapters and 4‑bit quantization, cutting GPU memory use by ~50 % while preserving output quality on NYU’s HPC cluster. • Curated a human‑feedback dataset and applied Reinforcement Learning with Human Feedback (RLHF) to steer model tone, accuracy, and compliance, boosting factual consistency by 10 % in blind evaluations. • Implemented a Model Context Protocol to register RAG search, real‑time FX pricing, and chart‑rendering functions as tools and invoke them through a single standardized interface. • Implemented MLOps: containerized the stack with Docker, automated CI/CD on GitHub, logged metrics to Weights & Biases.

Education

• Pohang University of Science and Technology

  Exchange

  2021 — 2021

• University of Electronic Science and Technology of China

  Bachelor of Engineering; Bachelor of Economics

  2018 — 2023

• New York University

  Master's degree

  2023 — 2025

• Hanyang University

  Exchange

  2022 — 2022