Sachin Gupta sachin1437

Hi, I'm Sachin Gupta 👋

👨‍💻 About Me

sachin = {
    "building"   : ["Mitra", "IERS", "SignBridge"],
    "focus"      : "AI systems that are deployed, not just demoed",
    "principle"  : "privacy-first, thought through before it ships",
    "stack"      : ["Python", "FastAPI", "React", "YOLOv8", "PyTorch", "Firebase"],
    "off_screen" : "Trekking in the Himalayas 🏔️",
    "reach_me"   : "linkedin.com/in/sachin-gupta-82909926b, sachingupta1437@gmail.com"
}

🔨 What I'm Building

🧠 Mitra, a privacy-first AI companion

An emotional support AI built around real privacy guarantees, not a privacy policy. Growing into a multilingual companion with an AR avatar in progress and a physical, hologram-capable form further down the roadmap.

Under the hood:

🔐 End-to-end encryption on conversation data
🚨 A dedicated crisis-detection layer, tuned and adversarially tested, not bolted on
🌐 FastAPI backend, React frontend, Firebase Auth and Firestore
🕶️ Incognito mode with a fully custom UI
🌍 Multilingual by design, not an afterthought

FastAPI React Firebase Gemini Three.js WebXR

(Mitra's repo is private, code link omitted, ping me if that changes)

🚨 IERS, Intelligent Emergency Response System

A multi-class threat detection platform built on YOLOv8, expanding from fire and smoke detection into weapons, people, and accident detection. A layered suppression system (temporal filtering, zone-based rules) keeps it from drowning operators in false alarms.

Shipped as a real product, not a notebook:

🖥️ FastAPI backend + React/Vite dashboard
💻 Packaged desktop app (PyWebView + PyInstaller)
🧱 Layered detection pipeline, built to scale to more threat classes

YOLOv8 FastAPI React OpenCV Docker

🤟 SignBridge, ISL Recognition

Real-time Indian Sign Language recognition using MediaPipe landmark extraction feeding a BiLSTM classifier, built to close a real accessibility gap for the hearing impaired.

Technical highlights:

🖐️ Custom-collected gesture dataset, multiple classes
🧠 MediaPipe hand landmark pipeline
⚡ Real-time inference via webcam feed
📄 Research write-up in progress for publication

MediaPipe PyTorch BiLSTM OpenCV

📋 Atmos, Face Recognition Attendance

Automated attendance for colleges using face recognition with liveness detection, built with a team of five, eliminating manual registers and spoofing risk together.

My contribution (ML + anti-spoofing):

👤 ArcFace-based face recognition pipeline
👁️ MediaPipe eye-aspect-ratio liveness check
🗄️ Django backend integration

Django ArcFace MediaPipe MySQL

🧭 What's Next

▸ AR avatar for Mitra (WebXR)
▸ IERS: multi-signal threat correlation
▸ SignBridge: wider gesture vocabulary
▸ More languages, more regions
▸ Open-sourcing pieces of the stack

🛠️ Tech Stack

Languages

ML / AI & Computer Vision

Frameworks & Deployment

📊 GitHub Stats

📜 Certifications

💛 Support This Work

If any of this is useful to you, or you just want to back independently-built, privacy-first AI, you can sponsor here:

🔗 Connect

"Ship it. Iterate. Repeat."

Navigation Menu