const akshay = {
  location   : "Thrissur, Kerala, India",
  education  : "B.Tech CSE (Cybersecurity) — Marwadi University, Rajkot, Gujarat · 2026–2030",
  building   : ["AI Second Brain (local-first)", "Fruvvi Studio", "SaaS MVPs", "AI Automation Pipelines"],
  engineering: ["Full Stack", "AI Agent", "Prompt", "Vibe", "Design"],
  security   : "Ethical Hacking · Web App Pentesting · CTF Practice",
  languages  : ["English", "Hindi", "Malayalam"],
  status     : "🟢 Available — internships · freelance · startup collabs",
};

I build production systems — not just projects.
Every line of code I write is designed to be deployed, used, and maintained.
Currently shipping a local-first AI operating workspace with evidence grounding, memory layers, and claim validation — from scratch.

A private AI workspace with auditable, evidence-grounded answers. No hallucination. No hidden magic.

This isn't a chatbot wrapper. It's a full AI operating system built on a custom truth architecture.

What it does:

• Ingests PDFs, Markdown, and text into a SQLite-first truth store
• Runs semantic + lexical retrieval over local documents (FAISS as rebuildable cache)
• Every answer is claim-validated — each factual statement is mapped back to the source chunk or flagged UNSUPPORTED
• 4-layer memory system: episodic, semantic, preference, and conflict memory with decay scoring
• Contradiction detection: conflicting memories are grouped, flagged, and surfaced — never silently blended
• Local plugin system with manifest-gated subprocess execution
• Full prompt firewall, API key scoping, and rate-limiting
• Streamlit UI with real-time evidence chips, retrieval trace, memory inspector, and audit views

Architecture snapshot:

PDF / Text / Markdown
        ↓
   Ingestion Layer  →  Document Profiling  →  Chunking
        ↓
   SQLite (source of truth)  →  FAISS (vector cache)
        ↓
   Query Firewall  →  Retrieval (semantic + lexical)
        ↓
   Reranker  →  Context Optimizer  →  LLM Synthesis
        ↓
   Claim Validator  →  Evidence Mapping  →  Contradiction Flags
        ↓
   Streamlit UI  →  Evidence Chips  →  Intelligence Panel

Current production score (internal audit):

Stack: Python SQLite FAISS Streamlit LangChain OpenAI API Docker Linux
Tags: #ai-agents #local-llm #rag #second-brain #privacy-first #production-system

Frontend & Full Stack

Backend & Data

AI & Automation

Infrastructure & Security

Design & Vibe Stack

  

B.Tech Computer Science Engineering — Cybersecurity Specialization
Marwadi University · Rajkot, Gujarat, India · 2026 – 2030
Focus: System Design · Network Security · Web Pentesting · AI Engineering

Let's build something real.