Skip to content
Nocta
[ 00:00:00 ]#b6d4c32 / EST. 2024

[ SOFTWARE & AI CONSULTANCY ]

We build intelligent software that ships.

Forecasting engines, decision tools, and automation — designed and built end-to-end, from first commit to production.

[ COMPANIES WE'VE WORKED WITH ]

  • Shell
  • Sarona Gems
  • Celtic Engineering

[ THE STACK BEHIND THE WORK ]

Next.jsTypeScriptReactPythonPyTorchTimesFMDuckDBPostgreSQLDockerOpenRouterPlaywrightSwiftNext.jsTypeScriptReactPythonPyTorchTimesFMDuckDBPostgreSQLDockerOpenRouterPlaywrightSwift

SELECTED WORK

Software we've shipped.

Five systems, built end-to-end. Every figure below is synthetic — we never publish a client's live data.

FEATUREDFEATUREDFEATUREDFEATURED
01
FuelCast
0 demand horizon, per grade
AI Forecasting · Retail Analytics/2026

FuelCast

A demand-forecasting and business-intelligence platform for fuel retail — predicting fuel and shop demand, catching margin leaks, and answering plain-English questions from operators.

PythonTimesFM 2.5DuckDBPlotly / Dash
View case study
02
Outreach Engine
0 compliance gates before send
Sales Automation · CRM/2026

Outreach Engine

A compliance-first outreach platform: source businesses, score their fit, draft the message, and keep a human firmly in the loop before anything sends.

PythonPostgreSQLPlaywrightLiteLLM / OpenRouter
View case study
03
Document Studio
0 median ingestion per document
Document AI · Workflow Automation/2026

Document Studio

A document studio that reads a request, drafts the reply, and signs the certificate — verbatim values where the source has them, grounded prose where it doesn't, a human in the loop before anything leaves the firm.

OpenRouterClaude (Anthropic)Tesseract / DocParserHandlebars
View case study
04
Sourcing Copilot
0 suppliers tracked in the directory
Supplier Discovery · Procurement/2026

Sourcing Copilot

A sourcing copilot that finds suppliers, scores their fit, and drafts the first enquiry — with a reviewer approving before anything leaves the desk.

OpenRouterClaude (Anthropic)PostgreSQLPlaywright
View case study
05
GUI Automation
0 element recall on legacy desktop apps
Computer Vision · Robotic Process Automation/2026

GUI Automation

A screen-aware RPA engine that scans a desktop, drafts a step-by-step script, and waits for a human to approve — then runs it, audited keystroke by keystroke.

YOLOv8 (Ultralytics)OpenRouterClaude (Anthropic)PostgreSQL
View case study

CAPABILITIES

Four ways we build.

From a first forecast to a shipped product — one team, the whole build.

01

AI & Machine Learning

Forecasting, computer vision, and LLM systems that make real decisions — not demos that die in a notebook.

  • Time-series & demand forecasting
  • LLM assistants & agents
  • Computer vision pipelines
  • Model evaluation & honest baselines
02

Data & Analytics

The pipes and dashboards that turn messy operational data into decisions your team actually trusts.

  • Data pipelines & warehousing
  • Interactive dashboards & BI
  • Reconciliation & data-quality gates
  • Metrics your operators believe
03

Automation & Ops

Quiet software that removes the repetitive work — integrations, internal tools, and workflows that run themselves.

  • Workflow & process automation
  • Third-party & API integrations
  • Internal tools & admin panels
  • Human-in-the-loop approvals
04

Product Engineering

Full-stack builds from first commit to production — web, mobile, and the cloud they run on.

  • Next.js & TypeScript web apps
  • iOS / mobile clients
  • Cloud deployment & DevOps
  • Design systems & UI

00 / METHOD

A short path from idea to production.

No six-month discovery phases. We prove it on your data fast, then engineer it properly.

01STEP

Scope

We pin down the decision your software has to improve, and the honest baseline to beat.

02STEP

Prototype

A working slice in weeks, not quarters — enough to prove the idea on your real data.

03STEP

Build

Production engineering: tested, reconciled, and reviewed adversarially before it ships.

04STEP

Deploy & measure

We ship it, wire the metrics, and keep iterating against what the numbers say.