Accidentally Awesome Labs live
LAB–LOG · EST. 2024
01 / 08
A design + development venture
making useful, delightful things for the web, the pocket, the community — and the occasional happy accident.
accidentally awesome LABS
Scroll

Awesome on purpose.

The team

Geometric illustrated portrait of Salar Sayyad

Salar Sayyad

Lead Designer · AI Developer · Solutions Architect

Founder of Accidentally Awesome Labs and of Genomarker. Designs and builds every product on this page, from interface to infrastructure, directing AI agents the way an architect runs a crew: spec first, gates always.

We design and build software end to end — the interface, the architecture and the hard parts underneath. The work that follows spans LLM fine-tuning, trade classification, gene-expression analysis, a native data workbench and a social archive.

AI agents bring the speed; specs, tests and hard gates decide what ships. If we claim it, there’s a receipt.

Capabilities

Product designDesign systemsWeb appsNative desktopML & GPU infraScientific computingAI-assisted engineeringAccessibility

Stack

TypeScript · Rust · Python · React / Next.js · Cloudflare Workers · Postgres · DuckDB · Axolotl / vLLM

The team

Geometric illustrated portrait of Salar Sayyad

Salar Sayyad

Lead Designer · AI Developer · Solutions Architect

Founder of Accidentally Awesome Labs and of Genomarker. Designs and builds every product on this page, from interface to infrastructure, directing AI agents the way an architect runs a crew: spec first, gates always.

Stack

TypeScript · Rust · Python · React / Next.js · Cloudflare Workers · Postgres · DuckDB · Axolotl / vLLM

Work 01· AI platform· 2026
Tessaro

Fine-tune a model. Get a verdict.

No-code LLM fine-tuning with a spend cap set before any GPU spins up — and a held-out verdict on whether it beat the base model.

AI / MLGPU infraCloudflareWeb app
tessaro.fm landing page with an example verdict card
W 1280 × H 960 · @2× CASE NO. AAL–003
Work 02· Trade intelligence· 2026
Harmony AI Client project

Classify products. Govern the AI.

A product classifier with text/image inputs, illustrative cost previews and bulk catalog workflows—backed by inspectable, governed AI.

AI / MLTrade complianceFull-stackModel operations
Harmony AI public classifier with a detailed industrial microcontroller description, image input, US destination and reporting-depth controls.
W 1280 × H 960 · @2× CASE NO. AAL–005
Work 03· Research platform· 2026
Genomarker Founder venture

Science that shows its work.

Gene-expression analysis that refuses broken study designs, with statistics validated against R and a signed receipt for every result.

BioinformaticsPython / FastAPINext.jsSigned provenance
genomarker.com landing page with an example assurance statement
W 1280 × H 960 · @2× CASE NO. AAL–001
Work 04· Native desktop app· 2026
dat0

There's life after row 1,048,576.

A native Rust + DuckDB workbench: open big files where they sit, record every transform, and seal the analysis into one replayable file.

RustDuckDBDesktopFile format
dat0.app landing page, built as the app's own workbench, showing the spreadsheet row limit
W 1280 × H 960 · @2× CASE NO. AAL–002
Work 05· Social app· 2026
MOGEN

No stars. Just how it felt.

A social archive for films, series and games: rate how it felt, keep a diary, find your taste twins. First commit to open beta in eight days.

SocialNext.jsPostgresPWA
mogen.fyi landing page with the 'How did it feel?' rating scale
W 1280 × H 960 · @2× CASE NO. AAL–004

Say hi.
Break things.

Write us

Follow along

Coming soon

  • Instagram
  • Dribbble
  • YouTube
© 2024–2026 Accidentally Awesome Labs · Made on purpose, mostly ← scroll back · we don’t mind the second look