AI data preparation and model fine-tuning
Entry Point AI
Entry Point AI helps developers and business teams prepare data and train task-specific language models. Its Transforms tool applies prompts across thousands of rows for tagging, extraction and classification. Synthetic Data expands existing examples into training datasets that users review and approve, while Fine-tuning turns selected data into custom models. The tools work independently or together, supporting tasks such as prioritizing support tickets, generating content and reranking search results.
Mark Hennings and Mihael Cacic started Entry Point AI together. Hennings described building the product in 2023 as a tool that uses AI to help train AI. Its engineering approach connects dataset preparation, prompt templates, training and evaluation through external model providers. A 2024 Llama-2 guide demonstrated that workflow without coding or configuring GPUs, including testing on held-out examples and downloading model weights through Replicate.
Entry Point sells subscriptions differentiated by pooled dataset rows and team seats, with the full toolkit included across plans. Fine-tuning and data generation run through customers’ own connected provider accounts, with provider charges separate from the subscription. Users can export their data as CSV or JSONL, including after cancellation.
1 talk
Newest first
1 speaker at AIE
Affiliations reflect their AIE appearances, not necessarily current employment.
Messages from the stage
From CSV examples to model testing
Hennings demonstrated importing CSV examples, generating missing press-release inputs, and combining GPT-3.5 Turbo fine-tuning with system prompts before testing the trained model in Entry Point Playground.
Affiliations reflect each recorded session, not necessarily current employment.