Kilani Sai Nikhil — Developer & Builder Portfolio

A modern, high-fidelity personal portfolio and resume dashboard showcasing machine learning tooling, full-stack applications, and computer vision systems.

Live URL: https://portfolio-t65m5dm5sa-uc.a.run.app

---

🚀 Technology Stack

• Core Framework: Next.js (TypeScript)
• Styling: Tailwind CSS
• Animation: Framer Motion
• 3D Graphics: Three.js & React Three Fiber (R3F)
• Icons: Lucide React
• Deployment: Google Cloud Run (via Google Cloud Build)
• Containerization: Docker & Docker Compose

---

✨ Key Features

1. High-Fidelity UI: Clean dark-mode aesthetic with vibrant accent colors, smooth glassmorphism, and responsive layouts.
2. Interactive 3D Elements: Background particle effects and interactive 3D structures powered by Three.js and custom GLSL shaders.
3. Interactive & ATS-Compliant Resume: Dual-mode resume page showing a rich dashboard UI on screen and rendering an ATS-optimized, print-ready page when exporting to PDF.
4. H-Fidelity Visualizations: Interactive system mockups, visual pipelines, and drone vision simulators for featured projects.

---

📁 Featured Projects

1. saara-ai (Local-First CLI & SDK Dataset Engine)

• Type: Open-Source CLI / SDK (published on PyPI and NPM)
• Tech Stack: Python, TypeScript, google-adk, crawl4ai, Ollama, vLLM
• Overview: Local-first CLI and SDK for high-fidelity dataset generation, labeling, and distillation. Utilizes google-adk to orchestrate autonomous ResearchAgents running firecrawl-local.

2. Vitt (On-Device AI Expense Tracker for MSMEs)

• Type: Full-Stack Mobile App (ready for Indus Appstore)
• Tech Stack: Flutter, Android AICore, Gemma 4 E2B, SQLite, Notification Listener
• Overview: Secure, local-first on-device AI financial tracking application for Indian MSMEs. Scrapes transaction notifications on-device (zero SMS/contact permissions) and schedules local Gemma 4 E2B models via Android AICore. Fully DPDP Act 2023 and SEBI AI Advisory compliant.

3. AerialEye (YOLOv11-Nano Aerial & Disaster Response Model)

• Type: Computer Vision Model (published on Hugging Face)
• Tech Stack: YOLOv11, SAHI Slicing, PyTorch, ONNX, TFLite (INT8), Dataset Curation
• Overview: Fine-tuned SUTRA object detection model specialized for high-altitude drone and disaster-response imagery (6 target classes: human, sos, vehicle, flood, road_damage, crack). Integrates Slicing Aided Hyper Inference (SAHI) and is quantized to INT8 to run efficiently on low-power Google Coral TPUs.

4. Super Orchestrator (Master Agent Orchestration Protocol)

• Type: AI Engineering Tool (published as Agent Skill)
• Tech Stack: TypeScript, GCP Vertex AI, Docker, Firecrawl, opencode CLI, code-review-graph
• Overview: Master agent orchestration protocol designed for token-efficient, cost-effective, and safe multi-agent execution. Implements a Task-to-Worker Routing Matrix (delegating lightweight tasks to free open-source models) and a Safelock Gatekeeper for risk-level classification to prevent prompt deadlocks.

---

🛠️ Local Development Setup

Option A: Standard Setup (Node & npm)

1. Install dependencies:

   npm install --legacy-peer-deps
   

2. Start the local development server:

   npm run dev
   

3. Open http://localhost:3000 to view it in your browser.

Option B: Dev Container (Docker Compose)

1. Build and run the local development container:

   docker-compose up -d
   

2. The dev environment automatically maps your workspace files and runs npm run dev inside the container.
3. Access the server at http://localhost:3000.

---

☁️ Cloud Run Deployment

Deployments are managed via Google Cloud CLI:

1. Ensure you are authenticated and have the correct project selected:

   gcloud config set project landing-pages-498606
   

2. Deploy the local folder (uses a .gcloudignore file to skip large directories and builds them in the cloud via Google Cloud Build):

   gcloud run deploy portfolio --source . --region us-central1 --quiet
   

3. Access your live service at the stable URL generated.