Aarush Gupta


My parents immigrated from north central India to the United States in the early 2000s. I was born in San Diego and moved around the bay before finding a home in Fremont, California.

I like living here; the town offers a community but leaves you alone with your thoughts enough to be your own person. I do not think I have become the person I want to be yet. I enjoy the proximity to San Francisco and Palo Alto, as well as the culture, music, and history of the area. There is something intoxicating about the acceleration of technology here and it makes me very excited for what the future holds.

I like the act of creative engineering and have been building for as long as I can remember. I do not know if I got into software and machine learning as a product of living in the Bay Area, or whether there was a genuine interest. I often imagine what I would pursue if I grew up in rural Missouri instead. I like to think I would be building in some medium regardless.

Like many, I believe the impetus of my work in startups was The Social Network. However, I think my general framework of good product design was developed at my apprenticeship at Google, where I built AI-enabled discussion over Google Classroom. The following summer, I did research at a stealth startup, where I worked on infinite context length in constant time for language models, which I attribute to my process of identifying and performing research.

Following that, I founded Silicon Valley Brain Company, a venture and r&d lab inspired by Bell Labs, where we worked on neuromodulation to make you superhuman, backed by Google AI. Our hypothesis was that it should be possible to manufacture focus by leveraging deep neural networks in conjunction with transducer arrays to stimulate specific neural cortexes. I took some time out to work on open research as well. I have since left the organization to build interesting futures, but do love the field, industry, and technology.

In the interest of working in high acceleration environments, I did some contract work at Clado (fka. Linkd), building out a beta of their search and discovery infrastructure from scratch. For the summer, I am working on Engineering at Artificial Societies in London, building Artificial Collective Intelligence.

Today, I spend my most of time building Aesara, working on information machines and chasing AGI. We have been building phenomenal systems and I cannot wait for you to see what I have seen.

I like to spend my free time on research: I lead a team at Cohere Labs working on multilingual language models, founded and run Celeritas Research, working on safe autonomous racing as a vector towards the rapid acceleration of self driving vehicles, and am an independent researcher with the John Hopkins Applied Physics Laboratory, working on emergent dynamics of multi-agent pursuit environments.

For fun, I love to help organize events and competitions. I like to lift heavy weights, read, travel, and talk to people much smarter than me. I spend a ton of time building side projects.

Looking to the future, I can only aspire to do great things. I have no clue what my life might look like 10 years from now, much less a week from today, and neither do I try to imagine or optimize for it. I find solace in destiny and that the heavens has a way of equalizing everything in the long run.

My Twitter and LinkedIn. For communication, I prefer [email protected].

an autodidact and factotum, believer in the american dream, and 17 years old.


plutus advanced open-source matching engine

The most advanced open-source matching engine for financial trading systems. Supports more trading scenarios than some production systems and includes a client library.
github.com/bxptr/plutus

TextbookGPT AI agents for AP textbooks

A collection of AI agents built from scratch for a set of popular high school AP textbooks, with infrastructure and MLOps from scratch. 1,000+ students and 24,000+ queries/month in ~1.5 months solely from word of mouth.
book.lipsti.cc

yavs Custom vector store for inference at the edge

Yet Another Vector Store. The only vector database that enables efficient insertion and retrieval cross-platform, within the browser with Web Assembly and locally with Python bindings.
github.com/bxptr/yavs

OpenVector agentic tree-based AI search

Agentic AI search for the masses. Uses a tree-based search algorithm alongside open-source LLMs for the fastest deep research system available, for everything from hard questions to vast topics.
openvector.xyz

Imperium AI summaries and analysis for SF legislation

Systematic and automated AI-based summaries and analysis of Board of Supervisors meetings from counties (currently only San Francisco County). Aims to distill and disseminate policies and legislation.
imperium.aarushgupta.com

Ghost Extension for autotyping for 10x productivity

A Chrome extension with the best machine learning models to emulate human typing to bypass revision checkers. Aquired a dataset and trained a robust and parameter-efficient model with customizable WPM.
ghost.lipsti.cc

Vita LLM humanizer

The most advanced GPT humanizer. Scientifically ensures text to bypass AI detectors (by working with probability and statistical theory) and outscores all major AI humanizers on detectors like GPTZero.
vita.lipsti.cc

LeGPT LeBron in a GPT

A fine-tuned LLaMA 3.3 over transcriptions of LeBron's various interviews and press conferences. Performs surprisingly good and drops a few gems now and then.
legpt.lipsti.cc

ClassFrame Collaborative AI-enabled work platform

A collaborative work platform and forum over Google Classroom. Supported all basic features alongside AI generated summaries over discussions and a grounded chatbot. Built at Google in summer 2023.
classfra.me

candle Train neural networks in the browser

A simple machine learning library to train neural networks in the browser. Made primarily as a joke to show that you can train a neural network in Firefox and still consume less memory than having more than a couple tabs open in Chrome.
github.com/bxptr/candle

murus Flask middleware for security

A production-ready middleware for Flask with common security and performance features. Includes JWT, mTLS, rate limiting, etc. and covers most attack vectors with an ergonomic API.
github.com/bxptr/murus

vektor Educational vector database implementation

A mini vector database implementation for educational purposes. Sub-linear time complexity on searches and locality-sensitive hashing for storage — fun and performant.
github.com/bxptr/vektor

sanity Language-dependent operating system in Rust

A language-dependent operating system in Rust, built to be the easiest system to add new architectures to. Implements all modern features expected from a stable system.
github.com/bxptr/sanity

oscar Efficient garbage collection for C

A really efficient, conservative mark-and-sweep garbage collector for C. Realistically, not for optimizing software but more so for beginners to learn the language.
github.com/bxptr/oscar

slurpee AI-powered link management system

An AI-powered link management system, based on embedding LLM generated summaries for webpages that are uploaded through a custom Safari shortcut.
github.com/bxptr/slurpee

Living in the Shadow of Death An investigation into the meaning of life in the framework of death.

The largest research and writing project I have ever done. A result of six months of work in AP English Language and Composition, LLM assisted.
/morality