RedditGPT
Fine-tuning (taking a model that is already trained and training it further on specific data) GPT-2, a text-generating language model, to generate and summarize content based on Reddit data.
Tennis Gym
Reinforcement learning for the tennis gym (gym here means a simulated environment where an agent can practice by trial and error) - Multi-DDPG (deep deterministic policy gradient) agents, which learn continuous control moves from rewards, with more than one agent learning at once
Reacher Gym
Reinforcement learning for the reacher gym - DDPG (deep deterministic policy gradient), an agent that learns to move a robot arm by trial and reward
Smart Banana
Reinforcement learning value based methods, which learn how good each action is rather than learning a policy (the rule an agent uses to pick which action to take) directly
AI Real Estate Agent
Developing AI-powered tools to optimize real estate search and engagement using RAG (retrieval-augmented generation, where the model first looks up relevant documents and then writes its answer with those documents in front of it), instruction fine-tuning (fine-tuning a model on examples that pair instructions with the desired responses, so it learns to follow instructions) with the OpenAI API (OpenAI's application programming interface, the way a program calls OpenAI's models from code).
Instruction-Tuning-LLM
Fine-tuning large language models for improved instruction following and domain-specific tasks.