kongzii

Ideated, built proof of concept, and shaped the development of a fraud protection agent for Safe & Gnosis App. The main contributor/designer of the Prediction Market Agent Tooling and Prediction Market Agents. Agents achieve 75% or above accuracy in the prediction of future events. Top agents have hundreds of dollars in profit, as seen on our Dune Dashboard.

Migrated in-house data pipeline into Databricks (running on Spark and Delta Tables), reducing the run-time of processing 60M emails from days to hours and halving the costs. Refactored 2 models to MLFlow and deployed them on Databricks’ serverless endpoints. Fine-tuned and served GPT3 models that write the best sales emails, increasing open and response rates. Created GPT-4 powered web scraper, navigating clients' websites and prefilling onboarding forms.

Delivered machine learning infrastructure based on Terraform (Terragrunt) infrastructure as code (IaC) on AWS. Deployed MLflow and Airflow to Kubernetes, including KEDA auto-scaling, XComs stored in S3, and workers configured for Fargate and EC2.

Increased the accuracy of the multilingual sentiment analysis (PyTorch) by 21%. Created a multi-modal model written in TensorFlow that reduces the workload of the other team by 84%. Optimised existing models to have more than 50% faster inference speed and lower memory usage. Implemented a reverse image and video search engine with PyTorch and FAISS, allowing quick search in thousands of images. Assured all experiments were 100% reproducible using tools like MLflow, Git, DVC, and Docker. Started a Python educational group, presenting tens of times about the latest Python topics.

Designed the architecture of the whole system, including the NLP and boosted trees model, for automatic product pairing, using Kafka for message streaming and Docker for deployment in Kubernetes, PyTorch and TensorFlow as deep learning frameworks, automating 80% of the traffic.

Built dozens of backends and frontends for marketing and reporting systems for BMW, GGTabak and others.