Client: Montreal-based AI company
Challenge: Our client’s projects almost always start with initial data sources that arrive in fairly messy formats. These data sources need to be cleaned up and structured. Once done, the data needs to be annotated in multi-level, nested formats, to achieve optimal input for the training process of their AI models. Currently, this can only mostly be done manually.
Solution: By deploying a 5-step Wrk Actions Wrkflow, our client was able to systematically clean up their initial input files (irrespective of format), and output structured classified results at a ~95% accuracy rate. More importantly, by parallelizing steps and filtering level 2 & 3 classifications Wrk Actions, the delivery speed was 200% faster than their regular in-house process.
See the full results of this case study here.