Epoch8 is a machine learning and computer vision engineering company based in Tbilisi, Georgia, with a team of more than 30 engineers. We build vision systems that run in factories, warehouses and stores, and we deliver them to the point where they work on the client's own hardware, on the client's own data, under the client's own operating conditions.
HOW WE WORK
Most of our projects start with a problem rather than a dataset. A typical engagement runs from a feasibility study on a small sample, through data collection and annotation design, model development and error analysis, to integration and a pilot on site. We stay through the pilot, because the first weeks in production are where a vision system either earns its place or reveals the assumption that was wrong.
The team is organised around that sequence. It includes computer vision engineers, data engineers, MLOps engineers and backend developers, so a single work package can cover data, models, serving and integration without subcontracting the parts in between.
WHAT WE BUILD
Visual quality control and defect detection. Surface defects, damage, deformation, contamination, missing or misplaced components, packaging compliance. We work on classes that are rare and expensive to collect, which is the normal case in industry.
Object and product identification. Barcode and label reading, OCR including embossed, engraved and low contrast text, and fine grained classification across catalogues of thousands of visually similar items.
Inspection stations and gates. Multi camera rigs that capture an object from several angles, reconcile the views into a single decision, and write one structured record per unit. This includes the optical design question that decides the outcome: where the cameras go, what lighting they need, and which angles can be dropped.
Retail shelf analytics. On shelf availability, planogram compliance and product recognition at store scale, including fresh categories where the appearance of an item changes from day to day.
Sorting and recycling vision. Product and material class recognition on moving conveyors, running on device.
Edge deployment. Models compiled and served on NVIDIA Jetson with TensorRT and DeepStream, tuned to a fixed latency and power budget.
Around all of it: annotation protocols, dataset and benchmark design, drift monitoring and scheduled retraining.
OPEN SOURCE
We maintain two open source projects and use them in client work.
Datapipe is a framework for incremental data processing and model retraining orchestration. It recomputes only what changed, which is what makes a pipeline affordable to rerun as new data arrives.
Vedana is retrieval infrastructure for question answering over documents and structured data, used where an answer has to be traced back to a source.
Both are permissively licensed and both are directly useful in a research project, where the difference between a demonstrator and a result is whether a partner can rerun the pipeline after the funding ends.
INDUSTRIES AND CLIENTS
We work with manufacturing, contract logistics, retail and FMCG, recycling and circular economy, robotics and agriculture. Clients we can name include Brickit, ACI and ECOBOT. Most other engagements are under NDA and we describe them by industry.
RESEARCH AND EUROPEAN COLLABORATION
We are looking to join Horizon Europe consortia as the partner that owns the machine learning or computer vision work package, and to carry it end to end: data collection protocol, annotation guidelines, dataset and benchmark release, model development, edge integration, pilot deployment at a partner site and the evaluation report. We can also take the reusability and exploitation tasks that surround a vision work package.
Georgia has been associated to Horizon Europe since 2021, so our legal entity participates and is funded on the same terms as an entity from a Member State. Georgia is also named in the WIDERA work programme among the associated countries with equivalent characteristics in terms of research and innovation performance, so our participation counts towards the widening and geographic balance of a consortium. Georgian organisations were in 58 Horizon Europe projects as of early 2025, so most consortia have never worked with one.
Our working language is English. We can