A record of some of the most memorable things I've personally done. I will try to continuously update this. This has turned more into a storyline of what I've done — oops!
2024
- Received support through the Masason Foundation
- Joined WAT.ai as a director of Education
- Helped build Globe Explorer's new image index — if the image search quality is better, thank me. I made a system to do over 10k caption, embeds, and stores a minute in return for a toy
- Did Youtube crawling infrastructure work — worked with proxy rotations and some cool threading hacks to get over 2000 high quality videos per hour
- Won the Milwaukee Bucks Hackathon by creating novel ticket pricing algorithms
- Trained an audio synchronization diffusion models
- Created a video interpolation hack with Stable Video Diffusion
Grinding on audio synchronization models for the past week — here's a result after some fine tuning and latent hacking
— Nima (@PourjafarNima) February 27, 2024
Send some other videos you'd want to see
Two main issues right now are audio quality and temporal consistency with super long videos — like the following pic.twitter.com/ecURRaxpuu
SVD can surprisingly do video upscaling decently
— Nima (@PourjafarNima) February 19, 2024
By noising all the latents in a video and passing them in as initial latents (much like vid2vid or img2img) — SVD does a decent job of making a more temporally consistent video
Left is the original, right is the new pic.twitter.com/24kkmHOBDf
2023
- Made a reinforcement learning agent to play Flappy Bird for my final project for my last year of high school computer science — shoutout to Ms. W!
- Started studying CS @ Waterloo
- Built AI tutors using RAG on textbooks at YR Hacks and won best AI hack
- Met Sam Altman — awesome person
- Helped out at the WOMBO office with RAG systems to embed all of archive and Twitter to fetch relevant papers/tweets on ongoing projects
- Eventually got hired to work there in the summer — I primarily worked on designing, training, and deploying video2video diffusion models, testing out pre-processing techniques for WOMBO ME, and also helping with full-stack work
- Here is some Vid2Vid content that went viral — we did 50 million views in less than a month across all creators
- Attended SPARC
- Won HackTheNorth 2023 with an AI wearable that combines video and audio — built this before any other AI wearables like Tab or Rewind came out. Was the first Toronto Manz to win Hack the North
- Received support from Tyler Cowen and Thiel Foundation through Emergent Ventures
- Took Waterloo's CS 145 course — >wrote about the Y combinator
- Built out infrastructure for prediction markets with CosmWASM and Rust to create real-time order books — worked on this with people from Soma Capital
- Started working as a Venture Scout with Soma Capital after that
2022
- Played with anonymous messaging using ZK SNARKS and compiling circuits in WASM — leveraged Semaphore
- Created Uniswap pool snipers — open sourced one of my scripts
- Did research with Peyman Momeni on multi-party computation and using it to prevent MEV on blockchains. Wrote various sample contracts such as a multi-call leveraging MPC and contributed to the core of FairBlock
- Started collaborating at Iron Fish on benchmarking different hash functions between bellman and Circom circuits
- Developed a "scratch-like" platform to make Solidity smart contracts — received a grant from Shaan Puri to make this and got a job offer from ThirdWeb to work in SF
- Started working alongside the amazing Andreas Veneris and Péter Garamvölgyi at Conflux — built cross-chain NFT bridges, optimized WebRTC packages in Rust, and wrote documentation about the network
- Gave a workshop on pandas @ York University
- Won ETH SF with Traverse — automated static analysis on all smart contracts deployed (I had college applications due the week of this, one of the most high adrenaline times of my life)
- Won the StarkNetHouse hackathon by creating a fully trustless EOA wallet recovery service
2021
- I used to run a financial literacy Youtube Channel. It was very cringe now that I look back on it — but we got to interview some really cool people (like James Leitner )
- Started working at Hack + — a nonprofit accelerator for student-led impacts. Built out their dashboard that manages $1.7 million in transactions. Learnt about how to make proper enterprise full stack apps (e2e tests, ci/cd, setting up monorepos, and also playing with GraphQL)
- Built an automated social media agent that won at RoboHacks
- A lot of hackathons — a bit too many....
- Helped found Send Story and my first encounters into "startups." Sole developer of Android and cross-platform version of the app. Collaborated with national athletes from team Canada, received user feedback which I iterated on, and designed/developed UI. Scaled to 3k DAUs in 15 different countries, and even had MRR
- Went to SHAD and voted for the best design project
- Finished in the top 3 of Canada for PicoCTF — focused on web exploitation
- Bought my first Ether and Doge — and started mining on a 1080 Ti
- Lots of math and computer science contests. I got onto the honour roll of some and also qualified for AIME. This is something I've done for a while, but first time I've seen some sort of noticeable results.
- Created an untraceable aim-bot for CS:GO using YOLOv5 and pyautogui — VAC doesn't detect programmatic inputs so this cheat is untraceable. Used by some friends in global elite.
2020
- Wrote my first Github commit
- Made crappy games in pygame — included tic tac toe, golf-simualtors, and even a starship shooter for my cs class
- Made data analysis tools for Smash Bros. Games using Slippi's open-sourced file formats
- Prototyped a 2D game engine in pygame
- Web-scraped data on all teachers in Ontario (salary, education, training, etc) from https://www.oct.ca/ — this later became a feature in an app I helped build
- Got my first internship! Worked at Diglit as a backend engineer — debugged Flask errors and learnt how to use Docker
- Scraped over 300k+ tweets and ran sentiment analysis to analyze Xenophobia during the pandemic for my politics course. Worked on this in conjunction with Erevena
- Took and sort of understood Andrew Ng's "Neural Networks and Deep Learning" course after learning about BERT. Due to having no linear algebra knowledge, I got completely lost in this course
- Participated and won my first hackathon! Worked on fine-tuning classification models to predict locust swarms given satellite imaging + gave crop planting suggestions via linear regression. This project ended up getting supported by Vanguard's ESG fund, the African Library Project, and even Microsoft — while trying to scale it to a village in Uganda
2019 (Grade 8)
- Started learning how to make websites with HTML and JS
- Ended up re-making my dad's business website — he paid me 10$ a month for this because that's how much he was paying before for it