Building Privacy-First AI Systems | Performance-Obsessed Full-Stack Engineer
"I architect local-first intelligent systems that prioritize data sovereignty, sub-10ms latency, and elegant developer experiences."
class SuyeshJadhav:
def __init__(self):
self.role = "Graduate Student @ NC State"
self.focus = ["Local-First AI", "Performance Engineering", "Agent Systems"]
self.current_work = [
"๐ญ TweetScape: NLP pipeline processing 500+ texts/sec",
"๐ค AI Assistant: 90 tokens/sec with <10ms retrieval",
"๐ฎ 3D Portfolio: Physics-based interactions with Three.js"
]
def daily_driver(self):
return {
"languages": ["Python", "TypeScript", "Go", "Dart"],
"ai_stack": ["Ollama (Llama 3/Gemma)", "ChromaDB", "Whisper", "RoBERTa"],
"backend": ["FastAPI", "Node. js", "PostgreSQL", "Redis"],
"frontend": ["React 19", "Next.js", "Flutter", "D3.js", "Three.js"],
"obsessions": ["Latency optimization", "Data sovereignty", "Clean architecture"]
}
๐ญ TweetScapeAgentic NLP Pipeline with Real-Time Visualization Challenge: Visualize polarization in social discourse
Solution: 4-model LLM pipeline + D3.js force simulation
Impact:
- 500+ texts/sec processing throughput
- Custom physics engine for sentiment clustering
- Real-time topic extraction & emotion mapping
Stack: FastAPI โข React โข D3.js โข Ollama โข RoBERTaKey Innovation: Custom ReAct agent orchestration with parallel model execution |
Privacy-First Agentic Framework Challenge: Build production-ready local AI system
Solution: Hybrid RAG + SQLite + ChromaDB
Metrics:
- 90 tokens/sec inference (Llama 3.2)
- <10ms vector retrieval latency
- Zero cloud dependencies
Stack: Python โข ChromaDB โข Whisper โข React NativeKey Innovation: Decorator-based tool registry with iterative reasoning loops |
๐ WolfCafe+ PlatformHigh-Concurrency Dining Infrastructure Challenge: Scale campus dining to 1000+ concurrent users
Solution: Optimized MERN stack + CI/CD pipelines
Performance:
- 450 req/sec sustained throughput
- ~2ms average API latency
- 91%+ test coverage (Jest/Supertest)
Stack: React โข Node.js โข MongoDB โข GitHub ActionsKey Achievement: Zero-downtime deployments with automated testing |
๐ฑ AcaMediaReal-Time Student Communication Hub Challenge: Bridge student-teacher communication gap
Solution: Cross-platform Flutter app with Firebase
Features:
- Real-time messaging & file sharing
- Role-based access control
- Offline-first architecture
Stack: Flutter โข Dart โข Firebase โข Cloud FirestoreKey Feature: Seamless offline-online synchronization |
const currentFocus = {
learning: [
"๐ฆ LangChain for advanced agent orchestration & tool chaining",
"๐ฆ Building a Redis clone in Go for distributed systems mastery",
"๐ฎ Advanced Three.js for physics-based 3D interactions"
],
building: [
"๐ค Multi-agent orchestration framework with tool chaining",
"๐จ Gamified 3D portfolio with interactive project showcases",
"๐ Real-time analytics dashboard for LLM performance monitoring"
],
exploring: [
"๐ฌ Quantization techniques for edge AI deployment",
"๐ WebGPU for client-side ML inference",
"๐ Homomorphic encryption for private AI computation"
]
};| Principle | Why It Matters |
|---|---|
| ๐ Privacy-First | Your data shouldn't leave your machine to be useful |
| โก Performance-Obsessed | Every millisecond of latency compounds at scale |
| ๐จ Developer Experience | Great tools should feel like magic, not work |
| ๐ Clean Architecture | Code is read 10x more than it's written |
| ๐งช Test-Driven | Confidence comes from coverage, not hope |
| ๐ต Flow State Soundtrack | ๐ฎ Favorite Games | ๐ฑ Currently Learning |
|---|---|---|
| Martin Garrix ๐ง | Stray ๐พ | LangChain ๐ฆ |
| Kygo ๐น | Valorant ๐ | System Design ๐๏ธ |
| Fred Again.. ๐ถ | League of Legends ๐ฏ | WebGPU ๐ฅ๏ธ |
I'm always interested in discussing:
- ๐ค Local-first AI architecture and privacy-preserving ML
- โก Performance optimization strategies for full-stack systems
- ๐จ Developer tooling and framework design
- ๐ฎ Interactive visualizations with D3.js/Three.js

