Skip to content
View ShulingTang's full-sized avatar
🎯
Focusing
🎯
Focusing
  • NUDT

Block or report ShulingTang

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
ShulingTang/README.md
Shuling Tang — Generative Vision

Research engineer in generative vision, five years in. Most of that time has gone into instruction-based image editing with diffusion models — getting them to understand what to change, touch nothing else, and run fast enough to ship. Off the clock, I apply the same principle to myself: anything an agent can do, I don't do twice.


Focus

Controllable image editing — instruction-driven local edits that preserve identity and layout; portraits, hair, and color are where "almost right" still looks wrong.

Distillation & acceleration — multi-teacher → single-student on-policy distillation; compressing dozens of sampling steps into a few without losing quality.

Evaluation — "did the edit work" can't be eyeballed; identity baselines, paired comparisons, and reproducible eval pipelines.

Agentic tooling — wiring training, image review, bookkeeping, and scraping into LLM agents. Built for myself, open-sourced along the way.


How I work

  • Baseline before claims — without a "leave it unchanged" control, every "improvement" is an illusion.
  • Data sets the ceiling: a week cleaning data beats a day tuning hyperparameters.
  • It's done when it reproduces locally; every automation runs on a smoke set first.
  • Writing down what broke is worth more than writing down what worked.

Toolbox

PyTorch · Diffusers · LoRA · FSDP / DeepSpeed · CUDA · Docker · Linux · Claude Code / MCP · n8n · Python


contribution graph

Less, but better.

Popular repositories Loading

  1. CodeofMachineLearningCourse CodeofMachineLearningCourse Public

    Experiment code displayed by the teacher assistant

    Python 2 2

  2. SDSC SDSC Public

    Scalable Deep Subspace Clustering

    Python 1 3

  3. MachineLearning MachineLearning Public

    Forked from ruthy-wei/MachineLearning

    Python 1

  4. KITTI-distance-estimation KITTI-distance-estimation Public

    Forked from harshilpatel312/KITTI-distance-estimation

    Estimating distance to objects in the scene using detection information

    Python

  5. TinyImageNet TinyImageNet Public

    Forked from fcipollone/TinyImageNet

    Course Project for cs231n

    Python

  6. CSCbox CSCbox Public

    MATLAB