Computer Vision & AI Development

Laan Labs designs and builds custom computer vision and AI systems and takes them from a first proof-of-concept to production software running on phones, tablets, headsets and servers. Laan Labs has been building software products for almost 20 years, many of them built on computer vision and machine learning, with more than 70 million app downloads, and works directly with clients as a founder-led engineering group rather than through layers of account management.

Most of the work is in the gap between a model that works in a notebook and a feature that works for a customer: collecting or generating the right training data, making the model fast enough to run on a device, handling the cases the demo never showed, and wiring the result into an existing product. That is where Laan Labs spends its time.

Projects range from a single model, such as real-time 3D pose estimation of cars for an automotive photography product, to complete pipelines that combine computer vision with large language models, such as CreatorView, which watches social media video and uses vision models and LLMs to assess creators for brand partnerships.

What we do

Consulting and feasibility

Honest assessment of what current computer vision and AI can do for a specific problem, what data it needs, and what it will cost to build, before anything is built.

Model development

Detection, tracking, pose estimation, segmentation, classification, measurement and 3D reconstruction, using fine-tuned foundation models where they fit and custom architectures where they do not.

Synthetic and real training data

Rendering synthetic datasets with automatic labels when annotated data is scarce, and building annotation tools when real data is required. Laan Labs has trained car and foot pose models this way at a fraction of manual annotation cost.

On-device deployment

Models optimized for Core ML, Metal and mobile GPUs so they run in real time on iPhone, iPad and Apple Vision Pro without a network connection, plus Android and edge hardware where needed.

LLM integration and fine-tuning

Language models combined with vision and product data: extraction, classification, agents and assistants, with fine-tuning and evaluation when prompting is not enough.

Integration into existing systems

Connecting new AI capability to the apps, APIs and back ends a client already runs, including improving or replacing an existing computer vision system.

How we work

A project starts with a short scoping conversation and, where useful, a few days of feasibility work on real data. The first deliverable is a working proof-of-concept, typically within weeks, that shows the approach on the client's own inputs rather than a benchmark.

From there Laan Labs builds the production version: the model, the deployment target, the integration and the tests, with the client's team involved throughout. Clients own the code and models that are delivered, and can continue with Laan Labs for ongoing development or take the work in-house.

Common questions