Work

Real projects with video demonstrations. Each includes technical writeups and production metrics from actual deployments.

Production-Ready AI Voice Agents

To deliver accurate, scalable FAQ call support, we built a custom AI voice infrastructure instead of relying on IVRs or black-box platforms.

<700ms Voice Latency — End-to-end response time across STT, LLM, and TTS.

Multi-Tenant Architecture — Isolated bots, sessions, and knowledge bases for each client.

Hallucination-Free FAQ Support — RAG-powered answers from verified knowledge sources only.

rag

Video demo

• Technical writeup

We Built an Offline RAG AI That Reads 1000s of technical Pages

We built a fully offline AI document intelligence system that reads thousands of pages of technical documentation and answers engineering queries instantly — no cloud, no internet, no hallucinations.

Exact cache hits — under 1ms

Semantic cache hits — ~10ms

26 second startup bottleneck reduced to under 1ms

computer vision

Video demo

• Technical writeup

AI-Powered Landmine Detection & Safe Route Planning System

We built an autonomous drone intelligence platform that generates aerial maps, detects potential landmines and hazardous objects using AI-powered computer vision, and computes safe navigation routes using pathfinding algorithms — enabling faster and safer terrain assessment in high-risk environments.

Generated high-resolution aerial maps from multiple drone video streams

Detected small hazardous objects reliably using YOLO + SAHI sliced inference

Improved detection accuracy on large stitched maps without downscaling

computer vision

Video demo

• Technical writeup

Smart Shot Analyzer – Basketball Trajectory Prediction & Collision Detection

Computer vision system that tracks & predicts basketball trajectory and detects hoop collisions using polynomial regression. Monitors R² drop below 0.99 to instantly classify rim hits versus successful shots.

Real-time ball tracking using OpenCV color filtering & contour detection

R² threshold (0.99) effectively distinguishes rim hits from clean shots

2nd-degree polynomial regression accurately models & predicts parabolic basketball trajectories

automation

Video demo

• Technical writeup

AI Task Scheduler: Intelligent Calendar Optimization with Agentic Negotiation & ML

AI-powered desktop scheduler merging OS algorithms, Agentic AI & XGBoost. Natural language task entry with real-time conversational adjustments.

91% accuracy in optimal algorithm prediction via XGBoost + SMOTE

65% faster task entry through NLP & automated validation

Sub-second chat responses for real-time schedule edits

automation

Video demo

• Technical writeup

Audio Denoiser: Speech-Noise Separation with U-Net

Deep learning audio denoiser using 2D U-Net on spectrograms to separate speech from ambient noise. Predicts Ideal Ratio Mask via MSE loss and reconstructs clean audio through inverse STFT.

U-Net architecture with skip connections preserves high-frequency speech details

Ideal Ratio Mask (IRM) prediction enables clean speech extraction from noisy spectrograms

Dynamic noise mixing at runtime eliminates need for pre-paired clean/noisy datasets

computer vision

Video demo

• Technical writeup

VidSage – Smart Video Transcoder & Analyzer

Automated video transcoding and quality analysis tool using FFmpeg and OpenCV. Transcodes to multiple resolutions, extracts stream metadata, and detects visual blurriness—all in a single Markdown report.

3 simultaneous resolutions transcoded in a single run (480p, 720p, 1080p)

Real-time blur detection using Laplacian variance across sampled frames

Automatic metadata extraction (codec, bitrate, resolution) via ffprobe

Loading work...