Bhavya Kansal – AI Engineer & DeepTech Innovator

ENGINEERING & AI

Hello, I am Bhavya Kansal

Founder – MultiModex AI

I architect and build scalable artificial intelligence systems, specializing in machine learning models, multimodal applications, and deep-tech solutions for real-world engineering challenges.

VIEW RESUME
1+
Years Exp.
10+
Projects
5+
AI Based projects Built
Bhavya Kansal Profile
OPEN TO ROLES

About Me

// Mission Statement

I build AI systems that actually ship — closing the gap between research-grade models and production infrastructure. Specialising in Deep Learning, Computer Vision, and Generative AI, I architect end-to-end intelligent pipelines that handle real-world complexity.

// Current Build

Founder of MultiModex AI — engineering multimodal reasoning infrastructure that unifies vision, language, and decision logic into a single deployable system. Creator of Project Aagni, an augmented generative AI assistant built for complex, real-time decision support.

RESEARCH
Driven
FOUNDER
Mindset
SYSTEMS
Thinker
AI & ML
PyTorch · TensorFlow · HuggingFace · Transformers
VISION
OpenCV · CNNs · YOLO · OCR · Depth Estimation
SOFTWARE
Python · C++ · JavaScript · REST APIs · FastAPI
HARDWARE
IoT · Robotics · Edge AI · Sensor Fusion · RPi

Current Focus

Active engineering domains & R&D areas I am building solutions for today

Generative AI

Designing and integrating LLM-based systems into real-world applications — structured prompting, API orchestration, and response validation pipelines. Building toward a fully Indian-built AI.

Active Research

Agentic AI

Building autonomous task-execution workflows using tool-calling, sequential reasoning chains, and backend automation scripts that run without human intervention.

In Development

Multimodal Vision

Cross-referencing logic across text, images, and sensor data to build systems that interpret the real world comprehensively — the core of MultiModex AI's architecture.

Core Focus

Hybrid Assistants

Combining LLM reasoning with deterministic logic layers to improve reliability and reduce hallucinations in deployed systems. Best of both worlds.

Prototyping

Ethical AI

Ensuring model safety, mitigating bias, and building transparent reasoning pathways for enterprise-grade and India-scale deployments. Responsibility by design.

Always On

Applied Automation

Directly embedding trained ML models into existing software architectures to automate redundant analytical pipelines — turning research into real ROI.

Production

CurrentExperience

AI/ML Intern Trainee
IIT Ropar

IIT Ropar

Indian Institute of Technology

Jan 2026 — Jul 2026 · 6 Months

Industrial AI/ML training program in collaboration with NIELIT, focusing on applied machine learning workflows, model experimentation, and real-world deployment pipelines.

ML Models Deep Learning Python Research
AI/ML Intern Trainee
NIELIT

NIELIT

National Institute of Electronics & IT

Jan 2026 — Jul 2026 · 6 Months

Structured industrial training program under Government of India's premier electronics and IT institute, covering AI foundations, model building, and applied software engineering.

AI Foundations Scikit-learn Data Science Govt. Certified

MYEducation

Upcoming

B.Tech — Artificial Intelligence & Machine Learning

2026 — 2029

Bachelor of Technology

TIET

Thapar Institute of Engineering & Technology

Patiala, Punjab

Advanced coursework in deep learning architectures, statistical modeling, and large-scale software engineering. Targeting specialization in AI systems and robotics.

Deep Learning AI Systems Robotics
In Progress

Diploma — Computer Science Engineering

2023 — 2026

Diploma in CSE · Final Semester

TPC

Thapar Polytechnic College

Patiala, Punjab

Built strong foundations in computer engineering principles, Python programming, data structures, and practical software development. Launching directly into B.Tech AI/ML at TIET.

Python DSA Java Final Sem

Engineering Projects

Applied implementations of machine learning models and software architectures.

NIELIT × IIT
Featured
Live on Render

Healthy AI

Heart Disease Prediction · Clinical ML · PDF Reports

A production-grade AI health prediction platform built during my internship training at NIELIT × IIT Ropar. Deploys a Gradient Boosting Classifier on clinical heart-disease datasets for real-time risk scoring, with interactive EDA visualizations and automated PDF report generation per patient. Full English UI, Flask backend, deployed on Render.

Python Flask Scikit-learn Pandas ReportLab Matplotlib
LIVE DEMO REPO
Live on Hugging Face

SignLang AI

Real-time ISL/ASL Translation · Computer Vision · MediaPipe

A real-time AI system that detects and translates Sign Language gestures into text and speech. Built using MediaPipe for precise hand landmark tracking and PyTorch LSTM for sequence classification. Engineered for high-framerate inference with a dark glassmorphism web interface.

Python MediaPipe PyTorch LSTM Flask HuggingFace
LIVE DEMO REPO
FAKE
83%
REAL
17%
Live on Hugging Face

DeepFake Scanner

AI Forgery Detection · Computer Vision · EfficientNet

A real-time AI-powered deepfake detection system built during my NIELIT × IIT Ropar internship. Uses EfficientNet-B4 fine-tuned on the DFDC dataset to classify media as real or AI-generated, with confidence scoring and visual forensic overlays. Deployed as an interactive web app on Hugging Face Spaces.

Python EfficientNet PyTorch Gradio HuggingFace
LIVE DEMO REPO
Telegram Powered

AAGNI Assistant

Conversational AI · NLP · Telegram Bot

Your Personal AI Assistant — AAGNI is an NLP-driven Telegram chatbot that handles real-time natural language querying, intelligent responses, and automated backend script execution. Built with the Telegram Bot API, featuring context-aware conversation management and live response streaming.

Python NLP Telegram API python-telegram-bot
VIEW CODE
Real-time CV · 89% Accuracy

Emotion Detector

Computer Vision · CNN · Real-time Inference

Trained and deployed a Convolutional Neural Network for real-time facial expression classification across 7 emotion classes. Achieved 89% validation accuracy on a 5,000+ image dataset. Uses OpenCV for live webcam feed processing with optimized inference pipeline.

TensorFlow Keras OpenCV CNNs
VIEW CODE
Facial Auth · Flask · WebSocket

NeuroLock AI

Facial Emotion Auth · Real-time Analytics · Anomaly Detection

An advanced real-time facial emotion recognition and classroom analytics platform — an upgraded, production-trained evolution of the Emotion Detector. Built with Flask + WebSockets + FER mini-XCEPTION, featuring live emotion tracking, session analytics, and ReportLab PDF report generation.

Flask OpenCV Socket.IO FER ReportLab
VIEW CODE
a3f9c2b1d8e4f7a0 · SHA256 · 0x4e2d...
VERIFY · 0xbc91a3 · BLOCK #4821 · IMMUTABLE
cert_hash=9d1f3e · VALID · OCR MATCH ✓
OCR · Blockchain · Cryptography

Immutable Doc-Verify

OCR · Simulated Blockchain · Document Integrity

Engineered a document verification pipeline combining Optical Character Recognition (OCR) with a simulated blockchain backend to guarantee digital certificate integrity. Extracts text via Tesseract, hashes it cryptographically, and cross-references against a tamper-proof ledger — ensuring documents are authentic and unmodified.

Python Tesseract OCR Cryptography SHA-256
VIEW CODE

Currently Open For

AI/ML Internships Research Collaborations DeepTech Projects Startup Partnerships
reach out

REACH OUT FOR COLLABORATION

Whether regarding technical roles, research collaboration, or discussing the engineering behind AI systems.

Email Me

Direct line for collaboration and professional inquiries.

kansalbhavya27@gmail.com

LinkedIn

Connect to discuss professional roles and networking.

CONNECT PROFILE

GitHub

Review my open-source models, codebases, and architectures.

VIEW REPOS