Sarvik Contact

/02

Work

Applied systems, algorithms, and technical prototypes, each with a detailed write-up. Built to hit a number, not to look busy.

  1. 01

    VLA-DSS

    A 28.9M-parameter, RGB-only Vision-Language-Action model for robot manipulation, built for the efficiency frontier: strong performance at a fraction of the parameters of generalist VLAs.

    • PyTorch
    • DINOv3
    • Wavelet scattering
    • Fourier Neural Operator
    • Robot learning
    // key result

    28.9M params, RGB-only, real-time on weak hardware, LIBERO Object 79.5%.
  2. 02

    Flow Matching for Robot Policies

    Integrating flow-matching action heads into the Octo generalist robot policy, generative action modeling for vision-language-action (VLA) systems.

    • Python
    • PyTorch
    • Flow matching
    • Robot learning
    // key result

    Generative action modeling for vision-language-action robot policies.
  3. 03

    Traffic Digital Twin

    A city-scale traffic digital twin: computer vision, graph neural networks, and reinforcement learning that watch real intersections, predict congestion, and optimize signal timings.

    • YOLOv8
    • PyTorch Geometric
    • Reinforcement learning
    • Streamlit
    • SQLAlchemy
    // key result

    Models Varanasi's Lanka → BHU Gate → Assi → Sigra corridor as a live, optimizable digital twin.
  4. 04

    HangOpt

    A dual-phase word-guessing solver: a greedy probability phase followed by a Minimum Volume Enclosing Ellipsoid (MVEE) optimization phase.

    • C++
    • Optimization
    • MVEE
    • Probability
  5. 05

    Axiom

    A high-performance 2D game engine built in modern C++, targeting 60 FPS with 10,000+ entities.

    • C++
    • OpenGL
    • SDL3
    • CMake
    // key result

    Engineered sprite batching achieving 166+ FPS by consolidating 1,000 sprites into a single draw call.

More on github.com/SarvikIIT, write-ups land here as projects mature.