adit magotra

18 • data sci @ aalto university, finland. (under construction; sorry for the ugly layout lol)

- phone: (+358) 041-741-6473
- github: alphalaser
- linkedin: in/aditmagotra
0. core @ aaltoes
- largest student-run entrepreneurship community in europe.
- leading build-it 2.0: shipping fast + more.
- partnering with killer startups on open-source.
1. president @ projectbeta (tech/startup club)
- increased members by 3x (37 → 120).
- organised intl comps with 500+ participants.
- oversaw work of 16 departments.
0. nlp research (independent)
- built a novel task detection dataset+model.
- presented paper @ nml workshop, neurips ‘23.
- selected for natl science fair (top 1% of 10k+).
- mailed mit phd for mentorship (she said yes).
1. swe/ml @ social nlp lab, smu singapore
- built a deployment interface from scratch for a p hot model based on the neural hawkes process.
- sent 200 mails to get this one lol.
- code, paper (naacl ‘22)
2. growth / community @ slingshot (techstars)
- onboarded 400+ high school tech prodigies.
- managed a community of 3k devs.
- helped scale to 10k+ active users and 200k arr.
- basically everything given a night with internet.
- 1st in data analysis+nlp comp.
- 2nd in biggest hs ml comp in india, app dev comp, cryptic hunt.
- 3rd in biggest mlthon in india + 2x cash prize.
- special award by symbl.ai, hacktable + 3rd in tranzathon. (hackathons)
- $1500 (full scholly) to attend stanford’s ulo: number theory.
- $1100 (full scholly) to attend linguistics camp @ ohio state.
0. beezy: google meet → todo list in real-time
- built dataset using linguistic filters and trained bert on it. teamed up with ex-apple engineer.
1. wolverine: self-healing rag agent
- a full feedback driven multi agent loop to detect its own hallucinations/misleading answers.
2. pine: globally hosted trie cli + rest api
- single global state allowing concurrent clients to run operations like autocomplete on a trie.
3. neural network from scratch
- built a neural network from scratch without any external libs in julialang.
4. gameplay images dataset
- wrote a python script to generate a dataset of gameplay images from videos on youtube.
- forgot about it → opened a few weeks ago to see it has 700 downloads.