Skip to content
View darshons's full-sized avatar

Highlights

  • Pro

Block or report darshons

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.

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
darshons/README.md

Darshon Singh

Electrical & Computer Engineering @ Rice University (B.S. 2028) · Houston, TX

Currently working on LLM-driven video generation at OpenStax (including LoRA fine-tuning of a locally hosted video model) and an advisor-matching system I built and sold to CareerSpring, a national nonprofit.

Darshon.Singh@rice.edu · LinkedIn


What I'm working on

Openstax-Undergrads · OpenStax, Rice University

A pipeline that turns a textbook section into a narrated animated video. An LLM plans the scenes, writes the Manim animation code, and repairs its own render errors in a feedback loop. I added a grid-anchor visual critic (ported from Code2Video) so the model can see when its own layout is broken, plus a per-scene router that picks the right rendering engine for each scene.

Python Manim FastAPI LLM codegen TTS

Advisor–advisee matching · CareerSpring (contract)

CSV ingestion, configurable constraints (matches per advisor, score thresholds, caps), Hungarian-algorithm optimal assignment, a review dashboard, and automated email-campaign integration. Scoped and sold it to the founder, shipped it solo, and I'm now extending it to match on unstructured text using embeddings.

Python Optimization Vector embeddings

Alpine-Club-Autofill

A Chrome extension built for a client running guided hikes in the Dolomites. Bulk-autofills Austrian Alpine Club membership forms from a CSV, including Korean address transliteration.

JavaScript Chrome Extensions


Selected work

MBAi / AgentFlow · 1st place, Claude Builder Club Hackathon

Built solo in 24 hours; won against teams of up to four. A drag-and-drop canvas where you wire up AI agents like an org chart and watch data flow through the hierarchy. Custom RAG pipeline with semantic chunking and a multi-turn tool-calling loop that lets agents spawn their own sub-workflows.

Next.js TypeScript Claude API RAG

elec378_final · Machine learning, image classification

Classical CV and deep learning compared head to head: SIFT bag-of-visual-words, patch-based sparse-coding dictionary learning, custom SVM kernels, ResNet with k-fold cross-validated hyperparameter tuning, and a VAE.

Python scikit-learn PyTorch Computer Vision

ta-chatbot · Rice Oshman Engineering Design Kitchen

A retrieval-based teaching assistant for the makerspace, so students can find equipment, machine protocols, and assignment guidance without waiting for a TA. Streaming responses, persistent conversation history, and a password-protected dashboard for reviewing what students ask.

Next.js TypeScript SQLite Ollama


Toolbox

Languages — Python · TypeScript · C · SQL

ML/AI — PyTorch · scikit-learn · LLM pipelines · RAG · vector embeddings · multi-agent orchestration

Systems — FastAPI · Next.js · SQLite · Docker · Git · Linux

Hardware — Embedded C · microcontrollers · sensor integration · rapid prototyping (CNC, laser, 3D printing)

Popular repositories Loading

  1. Openstax-Undergrads Openstax-Undergrads Public

    The undergraduate teams project for the Openstax summer 2026 internship

    Python 5

  2. MBAi MBAi Public

    TypeScript 1

  3. elec378_final elec378_final Public

    Jupyter Notebook 1

  4. English-Project English-Project Public

    Python

  5. Smart_OEDK Smart_OEDK Public

  6. ta-chatbot ta-chatbot Public

    Next.js teaching assistant chatbot with Ollama LLM integration, real-time streaming, and admin dashboard

    TypeScript