Skip to content

Commit e6df4cd

Browse files
committed
feat: initialize architecture-first README
1 parent 4bdbb0b commit e6df4cd

1 file changed

Lines changed: 56 additions & 0 deletions

File tree

README.md

Lines changed: 56 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,56 @@
1+
# Ayush Pandey // Architecture-First Portfolio
2+
**An Industrial-Grade Showcase for the "New Year, New You" Google AI Portfolio Challenge**
3+
4+
[![Deployed to Google Cloud Run](https://img.shields.io/badge/Deployed%20to-Google%20Cloud%20Run-blue?logo=google-cloud&logoColor=white)](https://ayushpandey.me)
5+
[![AI Engine: Gemini 3 Flash](https://img.shields.io/badge/AI%20Engine-Gemini%203%20Flash-orange?logo=google-gemini&logoColor=white)](https://aistudio.google.com/)
6+
[![Secured by Cloudflare](https://img.shields.io/badge/Secured_by-Cloudflare-F38020?style=for-the-badge&logo=cloudflare&logoColor=white)](https://www.cloudflare.com/waf/)
7+
8+
> **Production URL:** [https://ayushpandey.me](https://ayushpandey.me)
9+
> **Infrastructure Status:** Shielded via Cloudflare Edge (DDoS Mitigation)
10+
11+
---
12+
13+
## Engineering Vision: Beyond the Surface
14+
Most portfolios function as static galleries. This platform is an **Engineering Audit**.
15+
16+
I have architected this portfolio to expose the **"How"** and the **"Why"** behind my development ecosystem. Instead of a surface-level list, I provide a deep-dive into my flagship SaaS, **VersionCV**. Using **Gemini 3 Flash** to generate structured **React Flow** pipeline, I created a dynamic map of distributed service dependencies, allowing evaluators to audit the system design of my project in an interactive way.
17+
18+
---
19+
20+
## Portfolio Infrastructure & Security
21+
The portfolio itself is optimized for high-performance delivery while maintaining professional-grade security:
22+
23+
* **Shielded Origin Architecture:** The portfolio's **Google Cloud Run** instance is protected by a **Cloudflare Proxy**. This setup masks the origin IP, mitigates DDoS risks , and prevents "Denial of Wallet" attacks by filtering malicious traffic before it impacts the compute budget.
24+
---
25+
26+
## Google AI Integration: The Co-Architect
27+
28+
### 1. Gemini as the Intelligence Engine (Showcased Project: versionCV)
29+
**VersionCV** leverages **Gemini 3 Flash** for its core resume-to-job-description optimization engine.
30+
* **Hallucination Control:** I implemented a strict **XML Tagging** schema for prompts. This algorithmic approach enforces structured data outputs and ensures high-fidelity entity extraction.
31+
* **Performance Benchmarking:** R&D indicated that while various LLMs offer quality outputs, Gemini 3 Flash provided the **lowest latency** and highest **cost-economy** for high-token NLP tasks.
32+
33+
### 2. Gemini as the UI Architect (Portfolio Feature)
34+
I utilized Gemini to bridge the gap between abstract architecture and visual implementation within this portfolio. It assisted in generating the node/edge logic for the interactive **React Flow** pipelines that map out the internal services of the projects I build.
35+
36+
---
37+
38+
## The Cloud Run Optimization:
39+
This portfolio chronicles a real-world performance refactor on **Google Cloud Run**:
40+
41+
1. **The Bottleneck:** Initial monolithic deployments suffered from a **~5 second cold-start latency** when the service went idle.
42+
2. **The AI-Powered Audit:** Using Gemini to analyze Docker layer efficiency, I identified that unoptimized dependency overhead was bloating the image size.
43+
3. **The Solution (3-Stage CI/CD):** I refactored the deployment into a **Multi-Stage Docker Build**:
44+
* **Stage 1 (Deps):** Strict isolation of `package.json` to maximize layer caching.
45+
* **Stage 2 (Builder):** High-performance compilation environment.
46+
* **Stage 3 (Runner):** A minimalist **node:20-slim** production image.
47+
4. **The Outcome:** Drastic reduction in image pull times and optimized **Scale-to-Zero** configuration, resulting in significantly decreased cold-start times and zero idle costs.
48+
49+
---
50+
## Industrial Tech Stack
51+
* **AI/ML:** Google AI Studio (Gemini 3 Flash), Gemini CLI
52+
* **Infrastructure:** Google Cloud Run (Primary Compute), Cloudflare (WAF/DNS)
53+
* **Frontend:** Next.js (App Router), Tailwind CSS, Framer Motion, React Flow
54+
* **Runtimes:** Node.js
55+
---
56+

0 commit comments

Comments
 (0)