I'm a first-year Computer Science student at Lovely Professional University, building production-grade AI-integrated web applications independently, end to end: architecture, frontend, API integration, and deployment. My work centres on applied generative AI, wiring LLM APIs into real products rather than treating them as toys, alongside full-stack engineering with React, TypeScript, and modern tooling.
I work solo on every project in this profile, which means I own the full stack: design, state management, third-party API orchestration, deployment, and post-launch iteration.
🎯 Open To: AI/ML internships · Full-stack development roles · Open-source collaboration · Applied GenAI research exposure
Languages
Frontend
Backend & Data
Cloud, DevOps & Tooling
| Domain | Proficiency | Details |
|---|---|---|
| LLM API Integration | ⭐⭐⭐⭐☆ | Groq (Llama 3.1), production prompt chains, response streaming |
| Prompt Engineering | ⭐⭐⭐⭐☆ | Structured prompt systems for resume analysis, tutoring, and content generation |
| Clustering & Recommendation | ⭐⭐⭐☆☆ | K-Means genre segmentation, cosine similarity recommender (Spotify dataset) |
| Local LLM Deployment | ⭐⭐⭐☆☆ | Ollama (phi3:mini) served via Gradio for offline academic tutoring |
| Applied Generative AI Products | ⭐⭐⭐⭐☆ | End-to-end GenAI feature shipping inside deployed full-stack apps |
🚀 SpaceShield — No-Backend Space & Defence News Aggregator
India-first real-time aggregator pulling from NASA APOD, Spaceflight News, and GNews, with Groq-powered (Llama 3.1 8B Instant) summarization, fully client-side, zero backend infrastructure.
| Category | Detail |
|---|---|
| Stack | React, Vite, Tailwind CSS v3, Groq API, NASA APOD API, Spaceflight News API, GNews API |
| Scale | Multi-source real-time aggregation, no-backend architecture |
| Performance | Client-side fetch and caching, static Vercel edge deployment |
| Security | No server-side secrets exposure by design |
| Impact | India-first framing for space and defence news, underserved niche |
| Repository | github.com/Aaradhy-singh/spaceshield |
Architected as a fully static, serverless application to prove that meaningful multi-API aggregation doesn't require backend infrastructure, reducing hosting cost and attack surface simultaneously.
📄 ResumeAI Pro — AI-Powered Resume Diagnostic Platform
Browser-based resume diagnostic tool analyzing skill gaps, ATS compatibility, impact scoring, and role matching across 300+ roles. No account required, runs entirely client-side.
| Category | Detail |
|---|---|
| Stack | React, TypeScript, Vite, Tailwind CSS, shadcn/ui, jsPDF, PostHog, Formspree, Octokit |
| Scale | 8 analysis engines, skill gap matching against 300+ roles |
| Performance | Fully client-side analysis pipeline, PDF export via jsPDF |
| Security | No account required, no server-side data storage |
| Impact | End-to-end resume-to-hire pipeline: scoring, GitHub analysis, LinkedIn share, feedback loop via PostHog (6 custom events) |
| Repository | github.com/Aaradhy-singh/resume-ai-pro |
Built with a strict monochrome design system (black/white/navy, DM Serif Display + DM Mono, zero border-radius) to force disciplined UI decisions rather than defaulting to templated component libraries.
🎓 Local AI Academic Tutor
Offline-capable AI chatbot running a local LLM through Ollama, served via a Gradio interface, no external API dependency, no data leaving the device.
| Category | Detail |
|---|---|
| Stack | Python, Gradio, Ollama (phi3:mini) |
| Scale | Single-user local inference |
| Performance | Fully offline, latency bound by local hardware |
| Security | No external API calls, no data transmission |
| Impact | Privacy-preserving academic support tool with zero recurring API cost |
| Repository | github.com/Aaradhy-singh/local-llm-academic-tutor |
Chosen specifically to explore local inference tradeoffs versus cloud LLM APIs: cost, latency, and privacy in a single constrained project.
✨ Gesture Particle Universe — Real-Time 3D Gesture-Controlled Particle System
A single-file, 9,000-particle interactive 3D system controlled entirely via hand gestures, combining computer vision with custom shader rendering.
| Category | Detail |
|---|---|
| Stack | Three.js, MediaPipe, custom GLSL shaders, single-file HTML |
| Scale | 9,000 real-time rendered particles |
| Performance | Custom GLSL shaders for GPU-accelerated particle rendering |
| Security | Client-only, no data collection |
| Impact | Demonstrates real-time CV and WebGL integration in a zero-dependency deployable file |
| Repository | github.com/Aaradhy-singh/gesture-particle-universe |
Deployed as a single HTML file, a deliberate constraint to prove the entire CV and shader pipeline could ship without a build step.
No formal industry work experience yet. First-year student, actively evaluating legitimate AI/ML and full-stack internship opportunities through verified institutional channels only.
| Recognition | Details |
|---|---|
| Independent Shipping | 4 deployed, functioning full-stack/AI projects built and shipped solo |
| Certification Portfolio | Verified certifications across Oracle, Anthropic, AWS, Google, Microsoft, IBM, Hugging Face, JPMorgan Chase |
AWS
Oracle
Microsoft
IBM
Hugging Face
JPMorgan Chase
Anthropic
current_focus:
learning:
- "Advanced prompt engineering & LLM orchestration patterns"
- "System design fundamentals for scalable full-stack apps"
building:
- "Deployable AI-integrated products, solo, start to finish"
- "Portfolio depth ahead of B.Tech program start (Aug 2026)"
exploring:
- "Agentic workflows and tool-using LLM systems"
- "Local/offline LLM inference tradeoffs"
open_to:
- "AI/ML internships"
- "Full-stack development roles"
- "Open-source collaboration"
"Ship the deployed artifact, not the tutorial."