🤓 Latest Submissions
Treasury Agent
The Problem Federal grants are a massive source of non-dilutive capital, yet the path to securing them is paved with bureaucratic friction. Most existing AI solutions in this space are "chat wrappers"—static interfaces that generate text but lack operational output, stateful memory, or the ability to navigate live government infrastructure. The Solution fervoAI Treasury Agent is not a chatbot; it is autonomous infrastructure. It operates in a continuous stateful orchestration loop to bridge the gap between innovation and capital. Built for the Milan AI Week Hackathon, the agent identifies targets, calculates feasibility in real-time, and deploys specialized swarm workflows to close the execution gap. Key Technical Features Semantic Target Acquisition: Real-time query expansion against live Grants.gov S2S (System-to-System) API surfaces. Feasibility Matrix: Dynamic relevance ranking using Gemini 2.5 Flash across Technical Fit, Compliance Readiness, and Capital Efficiency. Dynamic State Injection: Operates with runtime awareness of team topology via fervo_state.json, ensuring tasks are assigned based on actual domain expertise. Swarm Orchestration: Decomposes complex grant applications into sub-tasks, routing them to human operators (Tech Lead, COO) and autonomous sub-agents. Impact By moving from "generative text" to "agentic orchestration," fervoAI Treasury Agent allows teams to focus on building while the infrastructure handles the hunt, the lock, and the deployment of federal opportunities.
19 May 2026
.png&w=828&q=75)
Startup Funding Architect
The Problem: Startups miss out on billions in non-dilutive federal funding every year—not because they aren't eligible, but because the process is broken. The search is fragmented, pitch writing is time-consuming, and deadlines are unforgiving. Founders need an autonomous partner, not just a search bar. The Solution: Startup Funding Architect is a production-grade Agentic AI solution built on IBM watsonx Orchestrate. It replaces the manual grant application lifecycle with a seamless, conversational workflow that moves from discovery to execution in under 60 seconds. Technical Architecture: We bypassed simple prototyping to build a resilient, enterprise-ready architecture: The Brain: An IBM watsonx Orchestrate Agent configured with a ReAct persona to reason through complex, multi-step user intents. The Muscle: A custom Microservices Context Protocol (MCP) Server built with Python 3.11 and FastAPI. The Intelligence: Integrated Google Gemini 1.5 (via the modern Google GenAI SDK) to power the "Triple-Horizon" pitch generation framework. The Infrastructure: Fully containerized (Docker) for reliability.Key Features: Resilient Grant Hunting: A custom scraping engine that queries the Grants.gov API with a 5x Exponential Backoff Retry Policy, ensuring data capture even during government server instability. Triple-Horizon Pitch Architecture: The agent doesn't just write text; it adheres to a strict prompt framework (Acute Pain Point, Technical Deviation, Strategic Lock) to maximize scoring potential against federal criteria. Proactive Execution: The agent anticipates user needs, automatically locking submission deadlines in Google Calendar and drafting emails in Gmail via secure, ephemeral OAuth token handling. Why This Matters: We didn't just build a chatbot; we built a workflow automator. By combining the orchestration power of watsonx with the speed of a hardened microservice, Startup Funding Architect proves that AI agents can be trusted with critical business operations.
23 Nov 2025
