Parallel 37 · Studio sheetEdition of 7 Oct 202637°00′N

We build AI products and grow them. Ours, and yours.

Parallel 37 is an independent studio. We build AI software under our own name, and we run AI growth for other companies: search, AI answers, ads, measurement and the site itself. Same tools, one rule: nothing goes on a page we cannot point to a source for.

Studio
Parallel 37 LLC
Filed
New Mexico, USA · July 2026
Entity ID
0008111584
Registered
1209 Mountain Road Pl NE, Albuquerque

Chart of works · the 37th parallel, Adana to Cupertino · not to scale

Flagships sit on index meridians in plate order · select one to open it

01The studio

Two kinds of work, one latitude.

AProducts we own

We build AI software under our own name.

Software for data teams, for clinics, for people choosing a perfume. The products are where the engineering gets tested: if a model can be wrong in a way that matters, a deterministic rule sits between it and the outcome.

BGrowth we run for others

We run AI growth for other companies.

Search, AI answers, ads, content, measurement and the site itself, from one accountable team. The growth practice is where the measurement habit comes from: measure first, write down what changed, refuse what cannot be shown.

The point

We are not a consulting firm with side projects, and not a startup pretending it has no clients. The products test the methods; the client work puts them to use.

02Flagships

Three products, one plate each.

Each plate says what the product does and exactly where it stands today, with a figure of how it works.

Plate I

MetricCourt

Beta Closed beta · pre-revenue
Status as of 7 Oct 2026

One business metric. One definition.

An AI agent for metric governance in companies that use dbt and the DataHub data catalog. It finds every model computing the same business metric, uses a deterministic SQL syntax-tree diff to show why the definitions differ, and writes one canonical definition back to the catalog.

Syntax-tree rules decide the verdict; the language model only writes the explanation. Run it with --no-llm and the verdicts are identical.

Verdicts
CONFORMING · DIVERGENT · UNKNOWN
Field test
6 public open-source dbt repositories, a 2,709‑model manifest, 15 hand-confirmed divergences, 0 crashesSource: parallel37llc.com
Tests
168 unit testsSource: parallel37llc.com
Licence
Core engine open source, Apache 2.0
Verdict readoutdbt → DataHub
metric
revenue
found in
3 models
orders.net_revenueCONFORMING
finance.revenue_monthlyDIVERGENT filter: status = 'paid' vs status in ('paid', 'refunded')
marketing.rev_dailyUNKNOWN definition could not be parsed; left for a person

canonical definitionwritten back to the catalog

Verdicts from syntax-tree rules · prose from the model

Fig. 1Illustration: how a verdict reads. Generic model names and SQL, not a customer run.

Plate II

Santralin

In development Not deployed · no customers
Status as of 7 Oct 2026

A reply desk that knows what it may not say.

An Instagram DM reply agent for clinics in Türkiye. It answers messages that patients start, inside the country's health-advertising rules, and turns each conversation into a callback lead for the clinic's staff.

It never gives prices, never diagnoses and never sends the first message. Those rules are not left to the model: a deterministic policy gate sits between the model and the send, running on a Cloudflare Worker.

Never
Gives prices · diagnoses · sends the first message
Built on
Cloudflare Worker, with a deterministic policy gate between model and send
Stage
In development. Not deployed, no customers.
Instagram DM threadGate on

Patient · first messageHi, how much does this treatment cost?

Model draftOur fee for this treatment is … TL.BLOCKED · PRICE

Policy gate
  • no prices
  • no diagnosis
  • never first

SentWe can't share prices by message. Can our team call you? What time suits you?

Callback leadto staff
Asked about
Treatment cost
Next step
Call back at the time the patient picks
Fig. 2Illustration: one message through the gate. Invented messages; no clinic, patient or account.

Plate III

KokuAsistan

Live kokuasistan.com
Status as of 7 Oct 2026

Perfume, read by what is in it.

A Turkish-language perfume discovery platform. Perfumes are described by their notes, their accords and the reviews people leave, so you can look for a scent by its composition instead of its advertising.

Catalogue
281,671 perfume recordsPer the live site, 7 Oct 2026
Language
Turkish
Built with
Next.js
Stage
Live. Anyone can use it today.
Note pyramid · accordsOne perfume
Top

bergamot · pink pepper

Heart

iris · rose

Base

vetiver · musk

Main accords
woody
powdery
citrus
musky
Fig. 3Illustration: how a perfume is read. A generic composition, not a listing from the site.

03Also on the record

Two smaller works, dated like the rest.

  1. Safelow

    LiveOn the App Store since 30 Jun 2026

    An iOS app of touch-reactive “floating ink” on calm water, rendered with Metal, with a paced-breath mode. Free, with a one-time purchase for more palettes. Nine languages. Published on the App Store by its developer.

    App Store

  2. Mijmara

    LiveLive since 17 Apr 2026

    “Mijmara — Gulf Fragrance Guide”: a bilingual, Arabic and English perfume discovery platform for the Gulf region, for finding Arab house perfumes and top Western luxury fragrances. Built with Next.js.

    Visit the site · mijmara.com

Live
Anyone can use it today.
Beta
Invited users only.
In development
Working code under test. Not deployed, no customers.

A status changes only when the evidence changes, and every status carries a date. The full register, with sources →

04AI growth

Growth for your company, the whole way round.

We run AI growth and marketing as a 360° loop: you get found, you get chosen, people convert, and every result is measured. What the measurement shows sets the next heading.

The growth loop: Found, then Chosen, then Converted, then Measured, which feeds back into Found. 000° Found 090° Chosen Converted 180° 270° Measured ONE PANEL
  1. 000°

    Found

    • Technical SEO and site audits. Your site, readable to Google and to AI systems.
    • Local SEO and Google Business Profile. Map visibility as a number on the panel.
    • AI-answer visibility. So ChatGPT, Gemini and Google's AI answers cite you.
  2. 090°

    Chosen

    • Content. Service pages, landing pages, articles. In healthcare, a regulation check first.
    • PR and digital authority. Publications that find you worth mentioning, not bought links.
  3. 180°

    Converted

    • Ads, Google and Meta. Every lira or dollar visible on the panel.
    • Web development, Next.js. Fast, measurable sites. We write the code.
  4. 270°

    Measured

    • Measurement setup. GA4, conversion tracking and a panel you can open. Work starts the day the counter runs.

Search and AI answers, ads, content, measurement and the site itself, from one accountable team, on one panel.

Refused

  • Ranking guaranteesNobody controls Google. A promised position is a number nobody plans to check.
  • Unlimited free workThe free part is the diagnostic.
  • Fake reviews or bent regulationRefused at any budget.
Request a free diagnostic

Site audit, AI-visibility scan, written report and a 45-minute call. The report is yours to keep.

05Contact

Tell us which half you came for.

baris@parallel37.agency

Pick a subject Products, investors, programs Growth for my company

A reply within 48 hours. No sales deck.