Alfred Tang Resume ↓

Alfred Tang

Computer Engineering at Georgia Tech, class of 2028. I build planners that merge autonomous cars into highway traffic in under 4 ms per decision, plus the occasional robotic gripper or AR game.

Open to internships: autonomy, robotics, ML.

Fig. 1 On-ramp merge. The ego car (@) predicts each gap with acceleration-aware extrapolation, plans a smooth merge path in station-lateral-time (dots), commits only if that path clears the predicted traffic, and steers along it. In the lane it holds a control-barrier headway (dashed box). Illustrative, written for this page. Tap or click a car to brake it.

Research

Two labs at Georgia Tech, one question: how does an autonomous car merge without hitting anyone, fast enough to run in the loop? Every number below comes from the work.

IVALab

Undergraduate Research Assistant

Prof. Patricio Vela

Jun 2026 – present

  • Planner benchmark harness

    A seeded evaluation harness that benchmarks 8 autonomous-driving planners on crash, merge, and runtime metrics. Batch runs fan out across 8 workers with ProcessPoolExecutor.

    serial5:00
    8 workers0:45

    Batch wall-clock time. 6.7× faster.

  • Real-time gap planner + CBF-QP safety filter

    Ported an offline MATLAB gap-based planner to a closed-loop Python controller for HighwayEnv, and added a CBF-QP safety filter that keeps lane-change trajectories physically feasible so merges stay continuous. It runs a 12 s space-time pipeline in under 4 ms per decision.

    MPC~7×
    gap planner<4 ms

    Time per decision, relative. About 7× faster than MPC.

  • Acceleration-aware behavior prediction

    Prototyped acceleration-aware behavior prediction for surrounding vehicles, measured across 500 paired episodes (p < 0.0001).

    crash rate−22%
    pred. error−55–74%
    merges/ep.+8%

    Relative to the baseline at 100.

  • One scenario format for two languages

    A shared JSON scenario format lets the MATLAB and Python planners run identical test cases. Running them side by side exposed a bug in the Python port that dropped valid merge options below the theoretical n+1 bound, which I fixed.

    scenario.json   (shape, illustrative)
    {
      "seed": …,
      "lanes": …,
      "ego":      { "x": …, "v": … },
      "vehicles": [ { "lane": …, "x": …, "v": …, "a": … } ]
    }
    MATLAB --+
              +--> same cases --> diff --> n+1 bound bug
    Python --+

ACT Lab VIP

Student Researcher

Autonomous and Connected Transport

Dec 2025 – present

  • PPO driving agent

    Trained a PPO reinforcement-learning agent for highway driving in HighwayEnv with Stable-Baselines3. The policy feeds the lab's CARLA integration and policy generation.

    step 09
    step 20k19

    Episode reward. Doubled in 20k timesteps.

  • Transformer fine-tuning

    Fine-tuned transformer language models for text comprehension with Hugging Face Transformers and PyTorch, and tuned the prompting and preprocessing pipelines around them.

    baseline1.0×
    fine-tuned~1.9×

    ROUGE, relative. About +90%.

Projects

Six builds from research, hackathons, and classes. Photos are drawn in text until you point at one; every card opens its write-up.

  • Space-time merge planner

    IVALab research · 2026

    Gap-based merge planning for autonomous highway driving, under 4 ms per decision, with a CBF-QP safety filter.

    • Python
    • MATLAB
    • HighwayEnv
  • 
                    
    In-headset view through Snap Spectacles: colored tetromino blocks stacked above a 5×5 grid playfield in a hallway

    Brain3D

    1st Place, Spatial Design · ImmerseGT · Apr 2026

    AR tetromino game for Snap Spectacles with hand tracking, surface placement, and real-time physics.

    • TypeScript
    • Lens Studio
    • Spectacles
  • 
                    
    Assembled gripper finger and servo housing wired to an Arduino Nano on a breadboard

    CV-controlled robotic gripper

    Feb – Apr 2026

    MediaPipe hand tracking at ~30 FPS drives an Arduino servo and a silicone-joint printed gripper.

    • Python
    • C++
    • MediaPipe
    • Arduino
  • 
                    
    Movies Store home page with navigation, hero banner, and footer

    Movies Store

    CS 2340, Objects and Design · Sep 2026 – present

    Django movie storefront, live on PythonAnywhere: search, cart, checkout, and review moderation that hides a review after 3 reports.

    • Python
    • Django
    • MySQL
  • 
                    
    Voxel chess board in space: the frosted white Unicorn, Ant, and Wizard face dark stone pieces; on hover a short clip plays the Unicorn capturing the Amazon

    C.H.E.S.S.

    Team project · Aug 2025 – May 2026

    Chess However Everything is Super Strange: Unity strategy chess where editor tooling defines 60+ pieces as data.

    • C#
    • Unity
    • Git
  • ML classification workflow

    Nov 2025 – Jan 2026

    scikit-learn pipeline that flags at-risk students; class weighting and a 0.35 threshold lift minority-class recall.

    • Python
    • scikit-learn
    • pandas
All project write-ups ->

About

Education

Georgia Institute of Technology

B.S. Computer Engineering, expected Spring 2028. GPA 3.92.

Coursework: Intro to AI, Data Structures and Algorithms, Design and Analysis of Algorithms, Object-Oriented Programming, Programming for Hardware/Software Systems.

Before the lab

Restaurant Assistant Manager, Saigon Cafe

Woodstock, GA · Jun 2021 – Aug 2025

Four years running a floor. I mentored and coordinated 4–5 employees through post-pandemic scaling and the return to full dine-in service.

Toolkit

Languages
Python, C++, C, C#, Java, JavaScript/TypeScript, MATLAB
Frameworks & libraries
PyTorch, Hugging Face Transformers, Stable-Baselines3, pandas, NumPy, Pytest
Tools
Git, Linux, Unity, Arduino, Jupyter, Google Colab, Claude Code, GitHub Actions

Contact

Hiring for autonomy, robotics, or ML? Email is fastest. The resume has everything on one page.