Nova Architect

• 

  Nova Architect UI

Inspiration

Cloud teams constantly get stuck between two bad options: overprovision for safety and waste money, or cut too much and risk outages. Existing tools can show alerts and recommendations, but they still leave people to figure out the tradeoffs, decide what to do, and manually execute the change. I wanted to build something closer to an infrastructure strategist: an agent that can understand constraints, reason across cost, uptime, and security, simulate the impact of proposed changes, and then safely help execute the next step. :contentReference[oaicite:0]{index=0} :contentReference[oaicite:1]{index=1}

What it does

Nova Architect is an autonomous infrastructure strategy agent built with Amazon Nova. It analyzes an infrastructure snapshot, identifies operational and architectural risks, proposes an optimization plan, explains tradeoffs in structured JSON, runs deterministic simulations for traffic spikes and failure scenarios, and then uses Nova Act to execute a safe, auditable demo workflow. It also generates an executive summary so the output is useful not only to engineers, but also to decision-makers. :contentReference[oaicite:2]{index=2} :contentReference[oaicite:3]{index=3}

The core idea is simple: move from static recommendations to reasoning, simulation, and safe action.

How I built it

I built Nova Architect as a full-stack prototype with a clean, judge-friendly interface and an agentic backend. The frontend is built in Next.js and presents a simple workflow: define the goal, analyze the environment, review the plan, compare before-and-after metrics, and optionally trigger a safe demo action. The backend is built with Python and FastAPI and orchestrates the reasoning flow. Amazon Bedrock serves as the unified model layer. :contentReference[oaicite:4]{index=4}

Amazon Nova is used in three distinct ways:

• Amazon Nova 2 Lite generates the structured infrastructure plan, including issues, evidence, recommended actions, tradeoff reasoning, and projected outcomes.
• Nova Act bridges planning to execution by automating a safe browser workflow and generating evidence artifacts for auditability.
• Amazon Nova 2 Sonic enables voice-triggered infrastructure strategy commands so the system can be used more naturally in operations workflows. :contentReference[oaicite:5]{index=5} :contentReference[oaicite:6]{index=6}

To keep the prototype stable and believable, I used strict schema contracts between the infrastructure snapshot and the Nova generated plan. I also validate the JSON output and retry once if formatting breaks. After the model produces a plan, I run deterministic simulations to compare estimated cost, uptime risk, and security posture before and after the proposed changes. :contentReference[oaicite:7]{index=7} :contentReference[oaicite:8]{index=8}

Challenges I ran into

The biggest challenge was making the system feel agentic and impressive without making it fragile. I did not want a vague chatbot or a fake dashboard. I wanted the output to be structured, measurable, and demo safe. That meant designing schema contracts, validating model output, keeping the simulation deterministic, and limiting Nova Act to safe workflows such as viewing configuration, tagging, or showing settings instead of performing destructive actions. :contentReference[oaicite:9]{index=9} :contentReference[oaicite:10]{index=10}

Another challenge was balancing ambition with reliability. For a hackathon submission, the project has to function consistently and match what is shown in the demo video. I therefore focused on a strong, production-inspired prototype rather than trying to build a fully live cloud optimization platform in a short time. :contentReference[oaicite:11]{index=11} :contentReference[oaicite:12]{index=12}

What I learned

This project taught me that the most interesting AI systems are not just generators, but decision systems. The real value comes from combining structured reasoning, simulation, explainability, and controlled action. I also learned how important trust is in infrastructure workflows: guardrails, recommendation-only modes, approval steps, and audit artifacts are what make automation usable in enterprise settings. :contentReference[oaicite:13]{index=13} :contentReference[oaicite:14]{index=14}

What's next

The next step is to connect Nova Architect to live telemetry, add a policy engine for approvals and governance, support staged rollout across dev, staging, and production, and add rollback aware remediation workflows. My long term vision is to turn this into a practical AI copilot for startups, SMBs, and lean ops teams that need stronger cloud decisions without a large SRE organization. :contentReference[oaicite:15]{index=15} Inspiration

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for Nova Architect

Built With

• amazonbedrock
• amazonnova2lite
• amazonnova2sonic
• amazonnovaact
• awsbudgets
• awsiam
• fastapi
• jsonschema
• nextjs
• python
• recharts
• tailwindcss
• typescript
• vercel