Argus

Argus Dashboard
Argus Landing Page

🛰️ Argus — Seeing Government Demand Before the RFP

Inspiration

Argus was inspired by how difficult it is for GovTech startups to detect government demand early. By the time RFPs are published, opportunities are crowded and relationship-building is already too late. Governments emit early signals through budgets, council minutes, and strategic plans, but these signals are fragmented and hard to interpret. Argus turns those hidden signals into a fast, visual, and actionable experience.

What It Does

Argus is an AI-powered government signal discovery platform centred around a real-time 3D globe. Signal hotspots appear across cities, allowing users to explore opportunities spatially instead of through static documents. Selecting a signal reveals context, budget, timeline, stakeholders, an AI-generated match explanation, and a draft outreach email.

How It Works

Argus uses a hybrid scoring system that balances semantic understanding with explicit requirement matching. Each startup is matched to government signals using the following weighted formula:

$$ \text{score} = 0.7 \cdot \text{semantic_similarity} + 0.3 \cdot \text{keyword_overlap} $$

Semantic similarity is computed using Gemini embeddings, capturing conceptual alignment beyond exact wording. Keyword overlap acts as a grounding mechanism, ensuring matches stay tied to concrete technical needs.

Architecture

The frontend is built with Next.js, Three.js, Tailwind CSS, and Framer Motion, delivering a performant cyber-style HUD and interactive globe. The backend is a single-file FastAPI service that integrates Gemini (gemini-2.5-flash) for match reasoning and outreach generation, using in-memory embeddings and pre-seeded signals for speed and reliability.

Challenges

Balancing real-time performance with AI-generated reasoning
Building a cinematic 3D interface without sacrificing clarity
Maintaining smooth globe interactions at 60fps
Managing variability in generative outputs during demos

What We Learned

How to combine embedding models with traditional scoring systems
How to build performant 3D interfaces in React under tight constraints
How much polish and visual clarity matter in demo-first products

What’s Next for Argus

Argus will integrate a more powerful Gemini model to improve semantic understanding, reasoning depth, and match explanations. This will enable more accurate signal interpretation and more tailored, reliable outreach generation.

Tech Stack

Frontend: Next.js, React, Three.js, Tailwind CSS, Framer Motion
Backend: FastAPI, Python, in-memory embeddings
AI / ML: Gemini (gemini-2.5-flash) for semantic embeddings and outreach generation
Databases / Storage: Pre-seeded signals (in-memory)
APIs / Integrations: Government open data sources, email API for draft outreach
Other Tools: GitHub for version control, Figma for UI design, Postman for API testing