AI-Powered Fleet Intelligence

Turn 2.3 million
vehicle records into decisions.

AutoIntel is an AI-native analytics platform built on the open Lithuanian transport registry. Forecast parts demand, drill into regional fleet trends, and query your market in plain language — all in one place.

Launch the app → See what's inside
✓ Real Lithuanian fleet data ✓ Gemini-powered AI ✓ Production-grade infra
app.autointel.lt/fleet-overview
Total fleet
2,333,665
↑ 1.2% YoY
Avg age
14.6y
↑ 0.3y
EV share
2.4%
↑ 0.8pp
Domestic origin
7.1%
→ stable
Monthly registrations · trailing 24 months
2.3M+
Vehicle records
9
Analytical surfaces
10
Apskritis regions
12mo
Demand horizon
Capabilities

Everything you need to read the market

Six pre-built dashboards, a full-fleet table view, AI-driven parts forecasting, and a chat agent that knows your data — purpose-built for the Lithuanian aftermarket.

Fleet Overview

Top-line KPIs, registration trends, and brand leaders at a glance. The pulse-check page for the entire 2.3M-vehicle dataset.

Brand & Model Analysis

Heatmaps, percentage-stack trends, and granular drilldowns across every marke × komercinis_pav combination on the road.

Regional Distribution

Apskritis-level maps, fleet density, and migration patterns — drill down to any of Lithuania's 10 administrative regions.

Fuel & Emissions

EV adoption curves, Euro-standard distribution, and fuel-mix shifts — track the pace of the transition vehicle by vehicle.

Age & Mileage

Histograms by production year, body type, and registration cohort. Asterisk-flagged provenance so derived values stay honest.

Import Origins

Domestic vs. EU vs. Non-EU bloc breakdowns, brand-by-origin matrix, and avg-age-by-origin to spot import-channel patterns.

Dashboards

Built like the platforms you already use

Single-graph-per-row layouts, theme-locked colors, tabular-numerals everywhere — every screen is built to the same design rubric so cross-comparing pages takes zero translation.

Fleet Database

Query 2.3M vehicles
in real time.

The classic spreadsheet view — but server-side filtered, paginated, and searched across marke + komercinis_pav with one keystroke.

  • Click-through faceted filters via the canonical FilterModal pattern
  • Multi-axis sort, paginated at the database edge
  • Fuzzy ILIKE across brand + model in milliseconds
  • URL-encoded filter state — share a query as a link
/api/v1/vehicles?brand=BMW&fuel_category=diesel&page=1
Brand
Model
Year
Count
BMW320d201812,408
BMW520d20179,317
BMWX520197,881
BMW318d20166,544
BMWX320185,990
BMW525d20154,472
Showing 1–6 of 142,803Page 1 / 23,801
AI Parts Forecast

Predict demand with
grounded confidence.

A two-stage architecture: Gemini supplies a cached service-schedule knowledge base; a deterministic SQL projection scores every vehicle in your fleet against that knowledge — never the other way around.

  • 3-, 6-, and 12-month demand windows pre-computed
  • 80% confidence intervals from cohort percentile bands
  • Mileage projection by age × fuel × region cohort
  • Admin review for low-confidence schedule rows
Brake pads · BMW 320d · 12-month demand
80% CI
2,140
3-mo p50
5,820
6-mo p50
12,340
12-mo p50
Live product

Production-grade
from day one.

Built on the modular-monolith pattern and deployed to Google Cloud — Cloud Run for the API, Cloud SQL for state, GCE for the worker stack, Secret Manager for everything sensitive.

  • Argon2id sessions, server-side, Postgres-backed
  • VPC-private database; no public surface
  • Sentry + Logfire wired end-to-end
  • Single-tenant, ready for multi-tenant the day you need it
AutoIntel login
AI Assistant

Ask your fleet
in plain language.

A Pydantic AI agent over Gemini Pro — grounded on your tables, with read-only SQL tools the model can call to fetch fresh numbers. Never invents data; cites every claim back to the row count it computed.

Try the chat → How it works
How many EVs registered in Vilnius county in 2024?
▸ AutoIntel In Vilnius county there were 3,847 EVs registered in 2024 — up +38% YoY from 2,790 in 2023. Tesla Model Y leads with 612 units (16%), followed by VW ID.4 (288).
Top 3 brands by 12-month brake-pad demand?
▸ AutoIntel Computing across 2.3M vehicles
Engineering

Modern stack, no shortcuts

Spec-driven, agent-implemented, ADR-tracked. Every architectural decision is committed to the repository as a versioned record.

FastAPI Postgres 16 Redis 7 Celery React 18 TanStack Router Chakra UI v3 Recharts Pydantic AI Vertex AI · Gemini Cloud Run Cloud SQL Terraform Sentry · Logfire
Ready when you are

Start exploring the
Lithuanian aftermarket.

Sign in to see live KPIs, drill into any region or brand, and put the AI assistant to work on your hardest demand questions.

Launch AutoIntel →