Growth Engine was a Kordspace internal concept/prototype focused on demographic personas using GPT-2. The project explored how a generative language model could support the creation of synthetic persona narratives from demographic context. Its scope centered on Python-based experimentation with model inputs and generated text, rather than a production customer-facing application. The concept positioned AI-generated personas as material for exploration and discussion, not as verified representations of real audiences.
The challenge
Demographic attributes alone cannot reliably explain an individual's needs, preferences, or behavior. A persona-generation concept therefore needed to distinguish useful fictional detail from unsupported assumptions about groups of people. The technical challenge was to explore controllable text generation while accounting for GPT-2's tendency to produce inconsistent, repetitive, or stereotyped content.
Our approach
The prototype's core approach paired Python experimentation with GPT-2 text generation. Demographic context served as the basis for prompts intended to guide synthetic persona narratives. A practical scope for this work included testing prompt structures and examining how generated responses varied in relevance, coherence, and specificity. Qualitative review was the appropriate lens for assessing those outputs, particularly where the model introduced traits not supported by the input. This framing kept the work focused on concept exploration rather than treating generated narratives as research findings.
The outcome
Growth Engine represented an internal exploration of GPT-2 for demographic persona generation. Its value lay in investigating a possible AI-assisted ideation workflow, rather than establishing audience accuracy or business impact. The concept also highlighted the importance of separating plausible language from evidence about actual people.
Technical highlights
- GPT-2-based synthetic persona generation
- Python-centered language-model experimentation
- Prompt-conditioned generation using demographic context
- Exploratory outputs distinct from validated audience data
Place in the journey
2018–2022 Logistics + AI + BlockchainGrowth Engine fits Kordspace's expansion into applied AI through internal experimentation with generative language models. It reflects an exploratory step toward broader AI and data-system work without implying production deployment or autonomous agent capabilities.