Engineering Director (Interim CTO) · Lightship Capital
AI Grant Writer on GPT-3 (2021)
Built a working AI Grant Writer in 2021 — well before ChatGPT — that enhanced grant response throughput 2.5x and delivered a 60% reduction in grant authoring labor.
- OpenAI GPT-3
- Semantic Search
- Prompt Engineering
- Python
- Web App
AI Grant Writer
In 2021, building anything serious on GPT-3 meant fighting context windows, instability, and a market that hadn’t decided generative AI was real yet.
This tool helped nonprofits and Lightship’s portfolio companies draft funding proposals against grant requirements they uploaded — needs statement, methodology, budget narrative, evaluation — in a conversational workflow grant writers actually wanted to use. Company knowledge and semantic search were unified to generate prompts the model could ground against, and the output became one of the early proof points that LLMs could do white-collar knowledge work, not just party tricks.
The numbers held: 2.5x throughput on grant responses, 60% less labor per grant.
The flywheel
This build became a lesson.
Every system Marlon ships turns into teachable patterns — in keynotes, workshops, and executive sessions. Want the lessons from this one, live?