Contact person: FPGA/HLS Expert Mentor for LLM Accelerator Project (HLSTransform)

Phone:Show

Email:Show

Location: Seoul, Korea, Republic of

Budget: Recommended by industry experts

Time to start: As soon as possible

Project description:
"Objective:
I am an engineer seeking an expert mentor to help me fully understand and successfully run an open-source LLM accelerator project on my local machine. The primary goal is to gain a deep, practical understanding of High-Level Synthesis (HLS) for developing FPGA-based AI accelerators.

This is a mentoring and guidance role, not a standard development contract.

Project Details:

Paper: "HLSTransform: Energy-Efficient Llama 2 Inference on FPGAs Via High Level Synthesis" (Cornell University, 2024)  

Open Source Code:([login to view URL])  

Freelancer's Role & Responsibilities:

Project Analysis & Explanation: Explain the overall architecture of the HLSTransform project, its C++ code structure (Host vs. Kernel), and the key HLS optimization techniques used (e.g., PIPELINE, UNROLL).

Guided Setup & Execution: Provide step-by-step remote guidance to set up the development environment (Vitis HLS, Vivado, etc.) on my computer, and troubleshoot any issues until the project compiles and runs successfully.

Technical Mentoring: Answer my questions about the core principles of HLS, the concepts of FPGA acceleration for AI, and how the code implements the ideas presented in the paper.

Required Skills:

Essential: Expert-level proficiency in the AMD/Xilinx Vitis HLS toolchain.

Essential: Strong experience in FPGA design and writing/optimizing HLS-compatible C++ code.

Highly Desirable: Familiarity with AI/ML concepts, particularly LLM architectures like Llama 2.

Crucial: Excellent communication and mentoring skills to explain complex technical concepts clearly.

I am looking for a knowledgeable partner to help me accelerate my learning in this specific domain. If you have the expertise, I look forward to your proposal." (client-provided description)