DEV Community: Abhishek Kumar

Turning Ordinary CCTV Cameras into Intelligent Safety Systems.

Abhishek Kumar — Sat, 17 Oct 2026 16:44:39 +0000

🚨 ANTAHDRISHTI (अंतर्दृष्टि): Turning Ordinary CCTV Cameras into Intelligent Safety Systems

🚀 Introducing ANTAHDRISHTI, an AI-powered public safety intelligence system designed to identify potential safety incidents using computer vision.

Traditional CCTV systems often rely on humans to continuously monitor footage. ANTAHDRISHTI explores how AI can help identify potentially dangerous situations and support faster responses.

💡 The Problem

In crowded public spaces, campuses, and commercial buildings, identifying safety incidents quickly can be challenging. Continuous manual monitoring is demanding, and important events may go unnoticed.

🧠 Our Solution

ANTAHDRISHTI uses computer vision to identify potential safety concerns through an interactive AI-powered dashboard.

✨ Key Features

🧍 Fall Detection
Identifies potential incidents involving a person falling.

🎒 Unattended Bag Detection
Helps identify bags that may have been left unattended.

👥 Crowd Detection
Identifies crowded areas that may require additional attention.

📊 Interactive Dashboard
Provides a centralized interface for viewing detection results.

🛠️ Tech Stack

🐍 Python
👁️ OpenCV
🤖 YOLO (Object Detection)
📊 Streamlit
🗄️ SQLite
✨ Gemini API (Optional)

🎯 Potential Applications

🏫 Educational institutions and campuses
🏢 Offices and commercial buildings
🛍️ Shopping malls
🚉 Railway stations and public spaces

The goal is to assist security personnel by highlighting potential incidents, not to replace human judgment or professional security teams.

🌱 Why This Project Matters

AI becomes more valuable when it addresses practical problems beyond generating text.

ANTAHDRISHTI explores how computer vision can contribute to public safety by helping identify situations that may require human attention.

This project reflects my interest in building practical AI applications with real-world potential.

🔗 Explore the Project

💻 GitHub Repository:
https://github.com/abhishek25001011-boop/AntaDrishti

I'd love to hear your feedback, suggestions, and ideas for improving AI-powered public safety systems! 🤝

🏷️ Tags

hf26challenge #devchallenge #opensource #AI #ComputerVision

PortfolioPilot AI — Describe Yourself. Get Your Portfolio.

Abhishek Kumar — Sun, 04 Oct 2026 14:39:13 +0000

This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend

What I Built

I built PortfolioPilot AI, an AI-powered portfolio generator designed for a friend who wanted a professional developer portfolio without spending hours writing content, designing sections, and putting everything together manually.

The idea is simple: my friend describes themselves in natural language, and PortfolioPilot AI turns that description into a structured, recruiter-ready portfolio.

It can generate sections such as:

About / Bio
Skills
Education
Projects
Achievements
Experience
Contact and social links

It also provides a live preview, editable sections, multiple templates, portfolio scoring, and AI-powered improvement suggestions.

Demo

The project is currently designed to run locally with Ollama for AI inference.

GitHub repository:
https://github.com/abhishek25001011-boop/PortfolioPilot-AI

The repository contains the complete frontend and FastAPI backend needed to run the project locally.

Code

GitHub:
https://github.com/abhishek25001011-boop/PortfolioPilot-AI

The main branch contains the stable version of the project.

How I Built It

PortfolioPilot AI is a full-stack application built with:

React + TypeScript + Vite
Python + FastAPI
Ollama
Qwen Coder, an open-weight AI model
LocalStorage for draft persistence

The core flow is:

User Prompt → FastAPI → Ollama → Qwen Coder → Structured JSON → Portfolio Preview

Instead of using a closed AI API, I used local inference through Ollama. The model generates structur bhaied portfolio information which the frontend then renders into the portfolio UI.

Why Does Open Innovation Matter?

Open innovation made this project possible to run with AI inference locally.

Using an open-weight model through Ollama means the core AI generation does not have to depend on a paid closed API. A user can run the model on their own machine and keep their portfolio information local.

It also makes the AI layer more flexible: the model can be changed or configured without redesigning the entire application.

For a personal tool like this, that gives users more control over their data, model choice, and running costs.

My Agent Session

Not included.

Prize Categories

No partner category selected.