Competed

Top Honours

Prize Money

Shipped

Credentials

class BalarajR:
    role       = "AI/ML Engineer | Full Stack Developer | Intelligent Systems Builder"
    location   = "Bengaluru, Karnataka, India 🇮🇳"
    education  = "PES University — B.Tech CSE (AI & ML), 2024–2028"
    contact    = "balarajr483@gmail.com | balaraj.vercel.app"

    domains = [
        "Agentic AI & Multi-Agent Orchestration",
        "Healthcare AI — HIPAA-aware, FHIR R4-compliant systems",
        "Agricultural AI & Smart Farming (rural India scale)",
        "Edge AI — On-device inference with quantized LLMs",
        "Cloud-Native Full Stack (GCP · Azure · Firebase)",
    ]

    building = {
        "VaidyaOS"   : "Offline-first Edge AI healthcare OS — runs LLMs on rural Android devices",
        "CareerLens" : "AI career intelligence — Google Gen AI Exchange National Winner 🏆",
        "AgriSence"  : "AI smart farming platform — Inferentia 2.0 State Winner 🥇",
        "HealthMesh" : "HIPAA-ready multi-agent clinical AI — Microsoft Imagine Cup",
    }

    mission = "Build ethical, explainable AI that empowers humans in critical decisions."
    values  = ["Offline-first", "Privacy by design", "Real-world impact", "Ship fast"]

What I'm actively shipping right now — updated regularly.

Problem: 65% of India's rural population lacks access to qualified doctors. Internet connectivity is intermittent. Clinical AI must work without the cloud.

VaidyaOS is an offline-first Edge AI operating system for healthcare accessibility. It runs quantized LLMs directly on Android devices — enabling clinical triage, symptom analysis, and multilingual medical guidance with zero internet dependency.

Stack: React Native · Gemma · Llama.cpp · Firebase · Python · FastAPI

Problem: Job seekers apply blindly — no feedback on why they're rejected by ATS systems or how to close their skill gaps fast.

CareerLens is an event-driven AI career platform that automates resume parsing, identifies ATS failures, and generates personalized upskilling roadmaps aligned to live job market demand. Selected as the National Winner of the Google Gen AI Exchange Hackathon.

Stack: Next.js 15 · TypeScript · Gemini 1.5 Pro · Cloud Functions · Firebase · React Flow

Problem: Indian farmers lose 20–40% of crops annually to preventable diseases and poor advisory access. Agronomic advice is expensive, English-only, and internet-dependent.

AgriSence is an AI precision farming platform that detects crop diseases via computer vision, delivers voice-assisted agronomic guidance in 7 regional Indian languages, and merges satellite NDVI data with weather models for predictive crop management.

Stack: Next.js 16 · Gemini 2.0 Flash · Firebase · Genkit · GCP · Python

OmniSence is a multimodal creative AI pipeline that transforms text prompts into narrated video stories. It chains LLMs → Imagen 4 → Cloud TTS → video stitching in a low-latency streaming architecture on Google Cloud Run.

• 🎥 Streaming Scene Generation: Async token-level interception pipelines rendered scenes in parallel
• 🖼️ Imagen 4 Integration: Text-to-image generation for each story beat
• ⚡ Containerized: FastAPI + Docker on Cloud Run — cold start < 1.2s
• Stack: FastAPI · React · Gemini 2.0 · Cloud Run · Imagen 4 · Docker

HealthMesh is a FHIR R4-compliant, HIPAA-ready clinical orchestration platform that coordinates 6 specialized AI agents for patient triage, medication safety checking, and prescription supply-chain routing.

• 🏥 6 Specialized Agents: Triage · Medication Safety · Interaction Checker · Pharmacy Router · Audit Logger · Patient Notifier
• 💊 EHR Bridge API: Maps FHIR resources to local pharmacy fulfillment networks
• 🔐 HIPAA Design: Secure Firestore rules, Azure Active Directory auth, audit trails
• Stack: Azure OpenAI · FHIR R4 · FastAPI · Firebase · Python

TaskForze is an agentic productivity platform where a supervisor AI node orchestrates 5 specialized sub-agents — coordinating task planning, automated testing, code reviews, deadline escalations via WhatsApp, and voice call alerts via Vapi.ai.

• 🤖 5-Agent Mesh: Supervisor → Planner · QA · Dev · Escalation · Analytics
• 📱 Omnichannel Alerts: Twilio SMS + WhatsApp + Vapi.ai voice calls for deadline escalation
• 🗄️ Semantic Search: AlloyDB + ScaNN vector index for intelligent task matching
• Stack: Google ADK · FastAPI · Cloud Run · AlloyDB · Twilio · Vapi.ai

CyberShield AI is a 5-layer security suite for SMEs featuring a multi-parameter vulnerability scanner, an LLM-powered threat categorization engine, and a low-level C-based offline mesh network for emergency encrypted communication.

• 🔐 5-Layer Audit: Network configs · Cloud IAM · Endpoints · Firewall rules · API exposure
• 📡 Offline Mesh: C-based AES-256 mesh topology for air-gapped encrypted relay
• 🧠 AI Threat Analysis: Gemini categorizes CVEs and generates prioritized remediation plans
• Stack: Python · C · Gemini · FastAPI · Vercel

Python · PyTorch · TensorFlow · Scikit-Learn · LangChain · LlamaIndex · Google ADK · Genkit · Gemini API · Vertex AI · Llama.cpp · GGUF Quantization · RAG · Multi-Agent Orchestration · Edge AI

Next.js · React · React Native · TypeScript · Tailwind CSS · FastAPI · Node.js · Django · REST · WebSockets · C · RISC-V

Google Cloud Platform · Microsoft Azure · AWS · Firebase · Cloud Run · Cloud Functions · Docker · Vercel · Netlify

PostgreSQL · Firestore · AlloyDB · ScaNN Vector Search · Redis · SQL

Building in public and contributing back to the ecosystem.

💡 Open to collaboration: If you're building in Healthcare AI, Edge AI, or AgriTech — let's talk. Reach me at balarajr483@gmail.com or balaraj.vercel.app

📫 Let's Build Something Together

"Build AI systems that support humans in critical decision-making — safely, transparently, and responsibly."

— Balaraj R  |  balaraj.vercel.app  |  Open to Collaborations 🤝