Anni Huang

@HuangAnni

LLM Training Operation Specialist @ ByteDance | Ex-Master of IT in Business (AI track) at SMU

SingaporeJoined July 2022

About

I’m Anni Huang, an AI researcher-in-training currently at ByteDance, specializing in LLM training operations with a coding focus. I bridge the gap between engineering execution and model performance, ensuring the quality, reliability, and timely delivery of large-scale training projects.

Available for

I’m open to collaboration in algorithm contests, academic research(RLHF), and AI innovation projects. My long-term goal is to drive impactful AI products—whether through research breakthroughs or vision-driven product management.

Anni Huang's blogs

HuangAnni's Bloghuanganni.hashnode.dev112 posts

Articles Comments1

Recently published

AHAnni Huanghuanganni.hashnode.devAug 14, 2025 · 1 min read

LLM for Coding Benchmarks and Datasets

Benchmark/DatasetYear CreatedInstitutionTime SpanDataset SizeData FormatLLM Testing Capability LiveCodeBench2024UC San Diego, Microsoft Research2024-ongoing (continuously updated)Several hundred problemsProgramming contest problems with test cases...

0

AHAnni Huanghuanganni.hashnode.devAug 14, 2025 · 2 min read

How to change a base model to a reasoning model?

Turning a base model into a reasoning model is essentially a post-training + data problem. The model’s architecture can stay the same — what changes is how it’s fine-tuned, what data it sees, and what training objectives you use. Here’s the typical p...

0

AHAnni Huanghuanganni.hashnode.devAug 13, 2025 · 2 min read

DeepSeek GRPO Explanation (Why we need it? How does it work? What are the findings?)

GRPO: Efficient RLHF via Relative Policy Optimization (Firstly introduced by DeepSeekMath, reference) Why GRPO? Problem with PPO: Slow, memory-intensive, and prone to reward overfitting in large-scale RLHF. GRPO’s Advantage: A compute-efficient ...

0

AHAnni Huanghuanganni.hashnode.devAug 13, 2025 · 2 min read

Beyond Pre-training: The Power of RLHF in LLM Alignment

Pre-training uses massive datasets and computational resources—often thousands of GPUs running for weeks or months—making it a domain dominated by top AI companies. Post-training is much lighter in cost and time (often days instead of months) and foc...

0

AHAnni Huanghuanganni.hashnode.devAug 5, 2025 · 4 min read

LeetCode Daily Challenge-3477. Fruits Into Baskets II(Easy, O(n²))

Problem Statement You are given two arrays fruits and baskets of equal length n. Place fruits into baskets following these rules: Each fruit type must be placed in the leftmost available basket with sufficient capacity Each basket holds only one typ...

0

Anni Huang

About

Available for

Anni Huang's blogs

Recently published

LLM for Coding Benchmarks and Datasets

How to change a base model to a reasoning model?

DeepSeek GRPO Explanation (Why we need it? How does it work? What are the findings?)

Beyond Pre-training: The Power of RLHF in LLM Alignment

LeetCode Daily Challenge-3477. Fruits Into Baskets II(Easy, O(n²))

Search Hashnode

Anni Huang

About

Available for

Anni Huang's blogs

Recently published

LLM for Coding Benchmarks and Datasets

How to change a base model to a reasoning model?

DeepSeek GRPO Explanation (Why we need it? How does it work? What are the findings?)

Beyond Pre-training: The Power of RLHF in LLM Alignment

LeetCode Daily Challenge-3477. Fruits Into Baskets II(Easy, O(n²))