Lusion AI is a training-data and annotation company started with one conviction: the highest-quality AI comes from the highest-quality human review of data, not shortcuts around it.
Founder & CEO, Lusion AI
Leads strategy, quality standards, and client operations at Lusion AI.
Mahir Shariyer founded Lusion AI to close a gap he saw firsthand: as computer vision, speech, and language models scaled globally, the quality of the human-reviewed data behind them had not kept pace. Too many annotation vendors optimized for speed and price alone, passing inconsistency straight into training pipelines.
Lusion AI was built around the opposite bet — that a disciplined, human-in-the-loop review process, run by a trained team and backed by transparent QA metrics, produces measurably better models. Based in Dhaka, Bangladesh, the company works with AI, robotics, healthcare, and autonomous-systems teams building products that depend on clean, well-labeled data.
Under Mahir's direction, Lusion AI has focused on three things from day one: rigorous three-layer quality assurance, honest project scoping instead of overselling capacity, and building a long-term annotation and data operations hub in Bangladesh.
Mahir works directly across every stage of delivery — scoping pilots with incoming clients, designing labeling schemas for new modalities, and reviewing QA metrics with the team — rather than treating leadership as a purely external-facing role. That hands-on involvement is why Lusion AI's proposals reflect what a project will actually take, not a generic estimate.
"Every model is only as good as the data it was shown. My job is to make sure the data Lusion AI delivers is something a client can trust without having to double-check it."
— Mahir Shariyer, Founder & CEO
The milestones that shaped how the company operates today.
Set out to build an annotation partner run on transparent QA and honest scoping, rather than the speed-and-price-only model common across the industry.
Built the annotator → reviewer → senior QA review chain that now runs under every project, regardless of modality or client size.
Grew the team's capability from image annotation into video, LiDAR, medical imaging, OCR, geospatial, and NLP — with dedicated workflows for each.
Leads a Dhaka-based team delivering annotation and QA for autonomous-driving, retail, and conversational-AI clients across time zones.
Illustrative image-annotation examples show the source image first. Hover, focus or tap a sample to reveal its annotation overlay.
These illustrative comparisons are not previews of confidential client datasets. See the separate video and LiDAR examples on our work page.
See all work examplesTo give AI, robotics, and healthcare teams of any size access to enterprise-grade annotation and dataset quality, without enterprise-only pricing or timelines.
To build Lusion AI into Bangladesh's most trusted AI data annotation and training-data company, and a recognized name for global AI teams sourcing quality-first annotation partners.
Every engagement runs on honest scoping, documented QA, and respect for the people doing the annotation work as much as the clients receiving it.
We never let a delivery deadline compromise the three-layer QA that catches inconsistency before it reaches your pipeline.
Our team is based in Dhaka and works across time zones for clients worldwide, combining local expertise with global delivery standards.
NDAs, access controls, and secure data handling are standard on every project — not an add-on for enterprise clients only.
We continually refine our tooling and QA process rather than treating annotation as a static, commoditized service.
Annotators and reviewers are trained, fairly compensated, and treated as the core of our quality — not a cost to minimize.
We aim for long-term relationships built on transparent metrics and honest capacity planning, not one-off transactional projects.
Tell us about your dataset and we'll discuss potential scope and next steps.