An internal AI workflow that reads incoming resumes, extracts structured candidate data, and helps HR review applicants against a role's requirements.

Python
LLM Workflows
NLP
Document Parsing
FastAPI

View case study →

0.0176 MS / FLOWSystem online

2.8M flows

Binary gate

Route

Attack class

Benign exits early · Suspicious traffic is classified

Research — IEEE-format paper

2,827,876 flows · 0.0176 ms inference per flow

Cascaded Network Intrusion Detection on CICIDS2017

A two-stage NIDS that routes 2.8M network flows through a binary detector into an attack-family classifier, at 0.0176 ms per flow.

Python
Scikit-learn
Gradient Boosting
1D-CNN + LSTM
Autoencoders

View case study →

≤ 100 MS / FRAMESystem online

Inference runs locally on Jetson Nano

Working prototype on device

≤ 100 ms inference per frame on Jetson Nano

Real-Time Smoke & Fire Detection on Jetson Nano

YOLO-based fire and smoke detection optimized with TensorRT to run at 100ms or less per frame on edge hardware.

YOLO
TensorRT
Jetson Nano
PyTorch
OpenCV

View case study →

REAL-TIME VIEWSystem online

One overhead feed resolves available spaces

Working prototype

Single overhead camera · Real-time occupancy view

Smart Parking Occupancy Detection

A computer-vision prototype that detects which parking spaces are free from a single overhead camera feed in real time.

YOLO
OpenCV
Python

View case study →

TWO INPUT PATHSSystem online

Live microphone and uploaded audio

Complete

Live microphone + uploaded audio input paths

Whisper Speech-to-Text Pipeline

A transcription system built on OpenAI Whisper, accepting both live microphone input and uploaded audio files.

Python
OpenAI Whisper
FFmpeg

View case study →