// signal acquired — hi, I'm

BASSAM

AI & Data Engineer — I build prediction engines that actually ship.

See the work ↓ Get in touch
LIVE — Goalio360 is predicting the World Cup right now
01

Selected
work

LIVE NOW PythonDuckDBMonte-Carlo SvelteKitKotlinCloudflare

Goalio360 ↗

A full prediction service for the 2026 World Cup, live right now. Scrapers feed a DuckDB lake; an Elo + Dixon-Coles match model — validated against 2018/2022 tournament holdouts — drives Monte-Carlo simulations with real FIFA tie-breaks and live knockout resolution. A build-free data path updates the site mid-match, and an AI match chat runs on carefully engineered context. Ships as a SvelteKit web app and a native Kotlin/Compose Android app — I designed the backend, the models, and both clients.

OPEN SOURCE LuaNeovimLLM AgentsContext Engineering

Hive.nvim ↗

A multi-agent harness inside Neovim. Role-based agents with tool isolation, parallel fire-and-forget subagents, expert advisors you can consult mid-task, and automatic context lifecycle management — nudge, compact, reset — so long autonomous coding sessions don't fall off a cliff. Define your own agents in plain Markdown.

IN THE LAB PythonR████████

Project: █████████

Large-scale statistical modeling of football managers. Years of data, very big traditional models, zero hype. That's all I can say for now — it declassifies after the World Cup final.

DECLASSIFIES IN --d --h --m --s

More experiments live on github.com/bassamsdata ↗

02

Who is
Bassam

I'm an AI & data engineer who ships end to end — from scrapers to data lakes to validated statistical models to the apps people actually open. My models are backtested before they're believed, my agents run overnight without me, and my pipelines survive match day.

0teams modeled
0languages — Py·Go·Lua·R
0native platforms
24/7live pipeline uptime
a.

Models that earn it

Nothing ships on vibes. Every model change must beat a leave-one-tournament-out backtest before it lands.

b.

Agents that run unattended

I build harnesses, not demos — context management, tool isolation, and orchestration that hold up over hours.

c.

Pipelines that survive match day

Live data paths decoupled from deploys, self-pacing loops, graceful degradation. The site never waits for a build.

03

Let's build
something

Have a model to validate, an agent to tame, or a pipeline that keeps falling over? I'm listening.

GitHub ↗