Leveraging my hobbyist passion for cybersecurity, I construct generative agents to streamline and enhance Aqua Security's current projects and workflows

• Architect, Build, and Scale a department to deliver next generation AI powered immersive storytelling • Talent Budgeting, Acquisition and Management • Generating innovations in structured NLG, Text Entity Extraction and Zero-Shot Transfer Learning • Goal based Planning and Motivations using Knowledge Graphs and guided Text Generation • Scaling OpenSource LLMs to production with MLOPS, GGML, and llama.cpp • Create and Operate automated data generation pipelines with humans-in-the-loop • Creating proof-of-technology demonstrations to acquire more capital • Deliver all of our research into the productionized game through crossfunctional collaboration with the whole game studio • Lead and Answer to all AI related discussion with investors • Text embedding, Emotional analysis, Structured Text Extraction, BERT, Blazingtext, HMMs, Meta Learning • Python, Pytorch, Tensorflow, Pandas, Numpy, Knowledge Graphs, Genetic Algorithms for Reinforcement Learning • Amazon Sagemaker, EC2, S3, Lambda, API Gateway

• Production grade LLM deployment with automated ingestion pipeline and on-line hyperparameter search • High throughput benchmarking on a custom reward model stack. • Use of vLLM, exLlama, GPTQ, AWQ to improve performance of models both in experimentation and in production. (Up to 1900 tokens per second on 13B model RTX4090) • Feature engineering of quantitative metrics such as RSI, MACD, Bollinger Bands, and EMA for financial modelling. Collect historical data, preprocess and generate normalized sequences of training data

• Deliver various Text/Voice chatbots on Dialogflow for use in client acquisition • Solutions and Product Architect on AWS and GCP stacks, empower and guide the team to deliver • R&D to increase/enhance our product offerings with custom machine learning • Directly carry out business development and client acquisition • Train junior developers to fill backend/serverless roles

• Created a custom neural network that utilized CNNs and Mel-Spectrograms to determine when the client has verbal oddities in speech. The classifier was trained on Sep-28k: A Dataset for Stuttering Event Detection from Podcasts with People Who Stutter. • Delivered the model and code to run live inference so that the client can identify how to improve their speech during meetings

• Hands on application of Pandas and AWS EC2, S3, and EMR to carry out cell-tower analytics for clients spanning over 300 million users • Delivered a robust ETL Pipeline to ingest data over multiple sources and produce development and planning insights for client EdotCo whilst adapting to changing requirements. • Spearhead R&D such as ML based enrichment of current data sources • Lead the team in adopting AGILE Software Engineering principles such as Test-Driven-Development, Continuous Integration and Deployment, Version Control, Peer Review and Pair Programming

• Experience training a Reinforcement Learning agent to control the compressors on an actual gas pipeline in South America. Solely responsible for ML implementation(DDPG, A2C, A3C) • Constructed a serverless microservice architecture to ingest training data and simultaneously train on multiple models to compare performance • Tensorflow, Keras, Golang, Python, Protobuf/GRPC

• Provide financial analytic systems to empower the local financial industry using Graph Database Technology, paired with Cutting Edge Machine Learning techniques. • Use of LSTMs to interface with graph based Knowledge Bases for the use of Natural Language Processing, for Question and Answer Retrieval • Produced reports comparing techniques in multiple research papers and how they can apply to our client- Agricultural Bank of China in dealing with credit-score estimation and fraud-ring detection. • Presented my findings to clients in Mandarin