🤓 Latest Submissions
Research Pilot
Research Pilot is an end-to-end AI research assistant built for the AI Factory hackathon. Give it a topic, and it autonomously searches the live web across multiple sources, then synthesizes the findings into a structured report: an executive summary, key findings organized by theme, points of disagreement between sources, and identified gaps in the research. What sets Research Pilot apart is its session memory. Every research run is stored in a vector database, so you can ask unlimited follow-up questions and get answers grounded in the same sources, without triggering a new search each time. This makes it feel less like a one-off search tool and more like a research analyst you can have a real conversation with. The frontend, authentication, database, and deployment were built using native.builder, giving the app a clean, production-ready interface backed by Supabase for auth and data. The actual research intelligence is powered by a custom FastAPI backend (deployed separately on Railway) that the native.builder app calls via API. This backend orchestrates: Groq for fast LLM reasoning and report synthesis (via LangChain) Tavily for live, multi-source web search Qdrant as a vector store, enabling grounded follow-up Q&A against a research session without re-searching The backend exposes two core endpoints, POST /research (topic to structured report plus session ID) and POST /followup (session ID plus question to grounded answer), cleanly separating the research and reasoning engine from the interface layer. Research Pilot demonstrates how a hackathon no-code/low-code frontend builder and a custom AI backend can be combined into a single cohesive product: fast to ship, but with real agentic depth under the hood
10 Aug 2026
