Agents · Growth · Motion
AGE 19

nineteen. agent engineer and growth lead. taught a pipeline to make launch films on its own, mined 2.04M messages for market signal, and helped route $15M+ through a Polymarket trading terminal.

Est. 2007, shipping since 2025
One file · No external requests
01Flagship case study · Growth lead

Growth lead at eighteen for the pro trading terminal on Polymarket, built by Enclave Money. $15M+ in volume routed, 37k unique visitors, 5,000+ traders, and 21 launch films that rebuilt the real UI in React, published on x.com/fireplacegg.

The unglamorous part too: analytics reconciled across four systems (PostHog, MongoDB, ClickHouse, Polymarket APIs), 1,000+ personalized outreach conversations with top traders, structured QA on releases, and sourced alt-data features, including the aircraft registry mapping behind a shipped flight-tracking feature.

Two of the 21 launch films: the terminal launch, and mobile

Volume routed
Launch films shipped
37k unique visitors5,000+ traders1,000+ outreach conversations4 analytics systems reconciled
02Case study · app.ghostmotion.ai


Lovable for Remotion: one prompt in, a finished product launch film out. The pipeline explores the real product in a cloud browser, storyboards it, rebuilds the UI as live React components, edits beat-synced to music, reviews its own frames, fixes its own defects, and renders the mp4.

Under the hood: parallel builder agents held together by interface contracts, a three-tier quality gate (motion lint, frame-diff scan, vision review), and ffmpeg motion metrics scored against professional reference reels.

app.ghostmotion.ai ↗
GhostShipper: a Ghostmotion launch film, one take, start to finish
Agent sessions in 10 days
Build time, first film to now
~30k lines of TypeScriptThree-tier quality gateRenders its own mp4
03Case study · Market research systems

01

Signal engine

Mined 2.04M Discord messages into evidence-weighted signals. LLM extraction plus embedding clustering.

02

News classifier

Wired straight into a trading API. Guarded live positions automatically the moment news broke.

03

LP rewards estimator

Models the order book to predict daily liquidity-program earnings. 3.2k lines of Python, 52 tests.

His analysis was cited 74x inside the private community where Polymarket's sharpest traders talk.

The order book the estimator models
Messages mined for signal
52 tests3.2k lines of PythonCited 74x by sharp traders
04The lab · Side projects

TweetLab

Will this tweet work, before you post it: a fine-tuned RoBERTa multi-head model plus per-metric regressors predict likes, retweets and replies from the draft. React front end, Python API on Google Cloud.

Frontend ↗Backend ↗

Black-Scholes for binaries

Five-minute crypto prediction markets priced as cash-or-nothing digitals: EWMA volatility, gated drift, fractional Kelly sizing, a backtest grid on live feeds.

Repo ↗

One Piece Showdown

Competitive battler balanced by a draft-budget economy instead of a type chart. Playable prototype.

50-character roster

clawd-alerts

WhatsApp alerting on EC2: LLM-gated news, whale trade tape, resolution watchers.

Always on

OSINT stack

ADS-B, Sentinel-2 and FIRMS triangulated for geopolitical forecasting. Fed a shipped product feature.

3 satellite + air feeds
05Capabilities
  • 01AI / agent engineering
  • 02Quantitative research
  • 03Motion design (Remotion)
  • 04Growth & analytics
  • 05Data engineering (MongoDB / ClickHouse)
  • 06Infrastructure (GCP / AWS / systemd)
06Distribution on X

Growth was not a slide in a deck here: I ran the distribution myself. Every launch film on this page premiered on @fireplacegg, a channel I wrote end to end, while the personal account documented the year in public. Numbers below come straight from the post archive.

7M+Xcombined impressions, two accounts
1.8MTOPimpressions on one post
10.5KFPfollowers on @fireplacegg
306 posts in 12 months 28.5K likes 9.7K bookmarks 74 posts past 10K impressions @aaravXBT: 1.5M impressions 2,499 followers there
07About

Finished school at seventeen. Took a gap year. Ran growth for a trading terminal at eighteen. Nineteen now, walking into college with a shipped portfolio.

Everything on this page was built end to end by one person: the agents, the data pipelines, the films, the growth loops. The pattern never changes. Find the loud problem, build the machine, let the machine ship.

2007Born
2024Finished school, at seventeen
2025Growth lead on a live trading terminal
2026Ghostmotion live · college begins
08Contact

For agents, growth, films, or market research systems, one address:

© 2026 Aarav Kataria Set in Archivo Black + Space Mono Built loud, shipped quiet