Ethan Ho

▸ Raised a pre-seed round ▸ Started as a way to help un-insured vehicles in South Africa be protected through a monthly coverage subscription model, without premiums/deductibles. Bootstrapped an MVP1 as a team of 2 @africaassure.com. ▸ We Piloted with 10 customers to validate the business model, but eventually lost traction but learned a tremendous amount about building something real.

▸ Utilized AlBert to create a chatbot that allowed engineers working in hardware verification to query the results of thousands of models with no code ▸ Model was fine-tuned on unstructured text data which was transformed through Structured JSON files with design verification model results ▸ Integrated Model and CLI into existing code-base and created a one-click deployment to Microsoft Azure

▸ Trained BERT using TensorFlow, on 200,000 scraped Reddit text entries and tested alongside a medically validated depression dataset to test the validity of training a chatbot using social media data. Goal was to create a chatbot to help detect levels of ~depression. Research was published at the PyCon2020. ▸ Utilized Spark and TensorFlow to automate Multiomic data pipelines to automate training and deployment of clinical data

▸ Used python to to clean, forecast (ARIMA) and output excel files to prototype a forecasting tool for ~30k unique sku's ▸ Wrapped pipelines in a flask web-app in order for business usability that contained graphs, forecasting tool-kit and model customizability.

Exchange Student