Full presentation and demo. EdgeLake integrated with AKAVE and Filecoin so edge nodes handle large image and video volumes for AI, without a central data store.
Roy Shadmon presented the EdgeLake integration with AKAVE and Filecoin: an end-to-end edge data platform that handles structured telemetry and large media at the same time, with no central cloud in the data path.
The use case demoed was a retail chain with stores in San Francisco and New York. Each store runs cameras feeding inference on people-counting. Each location deploys an EdgeLake operator node alongside its data sources. A master node and a query node coordinate metadata, and AKAVE plus Filecoin handle the storage of the actual image and video assets at scale. Behind a single SQL query, you can see counts across both stores without specifying which location holds which records.
For policy and access control, EdgeLake supports two equivalent modes. In the simpler mode, a master node acts as a registry that authenticates operator-node certificates and tracks policies in its ledger. Operator nodes poll for new policies periodically. In the more advanced mode, the same policies live on an actual blockchain (Ethereum smart contracts are the example given), and the master node is replaced by direct smart-contract reads. From the operator-node perspective, both modes feel identical.
The integration matters because it closes the loop on edge AI. Inference happens locally on the same hardware that holds the data. Structured outputs are queryable in real time across locations. The heavyweight media (training data, raw frames, archived video) lives on decentralized storage rather than on S3 or a single cloud bucket. The team is presenting an extended version of the demo at IBM TechXchange 2025 in Orlando.