Skip to content
View D-Vspec's full-sized avatar

Highlights

  • Pro

Block or report D-Vspec

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. Markdown is supported. 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
D-VSpec/README.md

Divyansh Verma

Software Engineer focused on systems programming, distributed systems, and ML infrastructure.

LinkedIn · GitHub · LeetCode · Blog · divi.verma1903@gmail.com


Projects

LLM Inference Engine — C++, GGUF

Repository · Blog Post

A GGUF-based LLM runner built from scratch. Custom tensor operations with multi-threaded matrix multiplication, memory-mapped file loading via Linux syscalls, byte-level GGUF parsing with custom 16-bit float structs, and RAII-based memory management for long-running processes.

Trustless RPC Proxy — Rust

Repository

A multi-provider JSON-RPC proxy with M-of-N consensus for data integrity. Includes a verification engine that walks Merkle-Patricia Tries and validates RLP-encoded proofs at the byte level, cryptographic state-root validation for Ethereum calls, and a real-time observability suite.

dvMeet — Next.js, Node.js, WebRTC, Socket.IO

Repository · Live Demo

A peer-to-peer video conferencing platform built on WebRTC for low-latency communication. Establishes direct P2P data channels using UDP hole punching, and handles complex NAT traversal via a self-deployed and configured STUN/TURN server (coturn).

dvBoy — C, SDL2, Assembly

Repository

A Game Boy emulator with cycle-accurate CPU and MMU components. Handles memory-mapped I/O and hardware synchronization, renders 8-bit graphics via SDL2, and debugs system-level errors down to the register and instruction level.


Publications

Agentic Multi-Hop RAG Pipelines (in progress) — Benchmarking Repository Research paper on implementing multi-hop RAG pipelines with agent-based retrieval.

Tokenization in LLMsBlog Implementing tokenization algorithms for LLMs from scratch in C++.


VIT Class of 2027 · Vellore, India

Pinned Loading

  1. dvMeet dvMeet Public

    a google meet clone in nextJS

    TypeScript 1

  2. dvBoyC dvBoyC Public

    A gameboy emulator made in C

    C

  3. find-me-a-breeze find-me-a-breeze Public

    JavaScript

  4. inferenceEngine inferenceEngine Public

    C++

  5. Terminally-Flapping Terminally-Flapping Public

    FLAPPY BIRD ON TERMINAL USING C++

    C++ 2

  6. dvSql dvSql Public

    A toy sql client

    C