Abhinay Karthik

Abhinay Karthik, full-stack and AI engineer

I build software that works in teams: multi-agent AI, full-stack products and the machine learning between them.

Studio lead at Orvantia

Hyderabad, 17.38°N 78.48°E, IST

About

Every dot on this page is an agent. Right now they are roaming. Keep scrolling and they take the shape of the systems I build: a loop, four specialists, three rings, a pulse.

I’m a B.Tech engineer from Neil Gogte Institute of Technology in Hyderabad. I lead Orvantia, a software studio of seven, and spend the rest of my time on multi-agent AI for healthcare and developer tools.

I like software built the way this page is: small, specialised parts that hand off cleanly and keep going when nobody is watching.

Projects shipped
7+
Technologies in use
15+
Domains
Web, AI, systems

Small parts.

Clean hand-offs.

Always running.

Four systems.
Four shapes.

Each one built from agents, models or both. Scroll through them: the swarm takes each one’s structure.

LiveFour specialists at the corners, one router at the heart: enrollment, safety, efficacy, general.

ClinicalAgent

A multi-agent platform for clinical trial intelligence: patient-to-trial matching, FDA safety analysis and streaming answers to clinical questions.

The hard part
Routing ambiguous clinical questions to the right specialist, reliably
What I built
An intent-classification layer that dispatches before the agent graph runs

LiveThree interlocking rings: the three doshas a plan has to balance.

AyurAhaar

AI nutrition built on Ayurvedic principles: personalised diet plans that a doctor approves before a patient ever sees them.

The hard part
Encoding centuries of Ayurvedic knowledge as a structured, queryable graph
What I built
A dosha, food and disease knowledge graph with embedding-based similarity search

In developmentHeartbeat traces receding in time: the wearable signal it reads before symptoms show.

EntraFlux

An AI safety platform for GLP-1 metabolic therapies: predicting side-effects from wearables, spotting counterfeit drugs, mining safety signals.

The hard part
Predicting subjective side-effects from objective biometric signals
What I built
PyTorch sequence models trained on HealthKit and Google Fit time series

Earlier builds

  • EventMaster

    Event booking with seat selection, QR tickets and an admin dashboard.

    Web app

    EventMaster landing page with the headline Elevate Your Events with EventMaster
  • Inventory System

    Stock, billing, supplier and customer analysis, with stock recommendations.

    Web app

    Inventory dashboard showing total products, low stock items and a product table
  • Lake Water Quality

    Upload a satellite image, and a CNN reads water-quality parameters from it.

    Machine learning

    Lake Water Quality Analysis upload screen with a satellite image of a lake
  • SafeStreets

    AI-detected road damage reports, repair management and a mobile app.

    ML and mobile

    SafeStreets admin dashboard with damage report counts and a recent reports table

What I reach for

Set by depth: the heavier the word, the more I have shipped with it.

Frontend

  • HTML, Expert
  • CSS, Advanced
  • React.js, Expert
  • Next.js, Intermediate
  • Tailwind CSS, Expert

Backend

  • Node.js, Advanced
  • Express.js, Advanced
  • Flask, Beginner
  • MongoDB, Advanced
  • PostgreSQL, Intermediate
  • Prisma, Intermediate

Machine learning

  • TensorFlow, Beginner
  • PyTorch, Beginner
  • Scikit-learn, Intermediate
  • Neural Networks, Intermediate

Languages

  • Java, Advanced
  • Python, Advanced
  • JavaScript, Advanced
  • TypeScript, Intermediate
  • C, Expert

How I got here

  1. 01

    The foundation

    C, Java and Python. Data structures, systems, and the discipline of writing code that holds up.

    C / Java / Python / DSA

  2. 02

    The web

    Full-stack engineering at Neil Gogte Institute of Technology. React, Node.js, real-time systems, and products like EventMaster and a live inventory platform.

    React / Node.js / MongoDB / Socket.io

  3. 03

    The agents

    Multi-agent AI and machine learning: clinical decision agents in LangGraph, graph-based nutrition intelligence, CNNs that read water quality from satellite images.

    LangGraph / LLMs / PyTorch / Graph ML

  4. 04

    The studio

    Orvantia. A team of seven in Hyderabad building automation, SaaS and custom software for growing businesses.

    Orvantia / Product / Delivery

The studio I lead

Orvantia

Software that runs itself. Orvantia is a software and automation studio in Hyderabad. We help small and growing businesses replace diaries, spreadsheets and WhatsApp follow-ups with automation, SaaS and custom software their teams actually use.

An orbit. We stay after launch, and the work keeps coming back around.
Based in
Hyderabad, India
Team
Seven people, hands-on
Abhinay's role
Studio lead
Site
orvantia.in

What we build

  • Business automation

    Systems that replace diaries, spreadsheets and WhatsApp follow-ups.

  • Custom business software

    Tools built around how your team works: tasks, inventory, attendance.

  • SaaS platforms

    Your own software product, with logins, roles and billing handled.

  • AI assistants

    Practical AI that sorts, summarises and answers, only where it saves time.

  • Dashboards and reports

    Sales, tasks and team performance at a glance.

  • Websites and web apps

    Fast sites and apps that work on any phone.

  • Mobile apps

    Android and iOS apps for customers or field teams.

  • 3D and premium websites

    Cinematic, WebGL-driven sites that feel alive.

  • Support and maintenance

    Fixes, updates and improvements after launch.

Client story, FactoryFlow

“Saving around 4 hours of my daily time.”
Senior Vice President, Prayagh Consumer Care, on FactoryFlow, built by Orvantia

Things we build for ourselves

  • NutritionOS

    Live, free

    A private nutrition and training tracker with a food database made for Indian kitchens.

  • ContextBridge

    Open source, MIT

    A resume safety gate for AI coding sessions.

  • EnteraFlux

    In research

    An AI health companion: cooperating agents for medication, symptoms, nutrition and progress.

  • Attendance by camera

    In demo

    Face-recognition attendance from cameras a business already has.

How we work

  1. 1

    Understand

    How the business runs today, and where time gets lost.

  2. 2

    Plan and quote

    Scope, timeline and price before any code.

  3. 3

    Design

    Screens a team understands on day one.

  4. 4

    Build and test

    Small visible steps, tested with real data.

  5. 5

    Launch and support

    Set up, train the team, stay on afterwards.