These examples show the type of work Luniqex can deliver without disclosing client-sensitive information. The focus is the method: data quality, workflow fit, evaluation, security, and adoption.
Luniqex developed a proof of concept for estimating grass and fine-fuel curing percentage from ordinary field images. The work evolved from an initial vision experiment into a curated image-regression system using DINOv3, VLM-assisted review, label auditing, and month/season/region calibration.
The best grouped benchmark improved from 15.32 MAE in the early DINOv3 model to 7.85 MAE after dataset improvements and calibration.
A local AI dictation workflow was built to reduce outsourced processing and cyber risk. The system integrated multiple speech-to-text engines, local language models, and retrieval-augmented generation to support a seamless internal pipeline without sending sensitive content to external AI services.
The result was a controlled workflow where transcription, structuring, and drafting could happen locally while preserving data boundaries.
A practical automation concept for operations teams: collect recurring inputs from spreadsheets, email exports, documents, and system reports, then turn them into a structured review pack with flagged exceptions and draft follow-up actions.
This kind of engagement focuses on reducing manual collation while keeping human review and business accountability in the loop.