Hi, I’m Neeraj.

Computer Science Engineer specializing in AI, Machine Learning, and Deep Learning. I design end-to-end intelligent systems—ranging from automated visual models to data-driven pipelines—that transform complex multi-modal inputs into practical solutions. Passionate about bringing production-ready artificial intelligence to real-world applications.

See the work Email me

[02]

Skills

A snake that eats them. It plays itself until you take over with the arrow keys or a swipe; the full list is written out underneath.

autoplaying — click or tap to steer

length 3collected 0/16

used in a projectcurrently learning

Python
Java
Swift
Kotlin
MySQL
Django
AWS
Git
AI
Machine Learning
Deep Learning
Computer Vision
RAG
Vector Search
LLMs
NLP

[03]

Selected work

Efficient ANPR

Jan 2026Prototype

Reads plates off traffic video by running OCR exactly once per vehicle instead of once per frame — the tracker, not the detector, is what makes the pipeline affordable. The plate detector reaches 98.3% mAP@50; the tracker is what decides how often it has to matter.

Python, YOLOv8n, DeepSORT, PaddleOCR, OpenCV

Code ↗

PolicyMind AI

Aug 2026Prototype

Answers questions across eight company policy PDFs — 80 pages — and cites the document and page behind every claim. The part worth pointing at is not that it answers: it is that a relevance floor lets it return nothing, so “the documents don’t specify that” is a real output rather than a failure mode.

Python, ChromaDB, Gemini 2.5 Flash, sentence-transformers, BM25

Code ↗

PingDrop

Mar 2026In development

Notifications, clipboard, and files move between an Android phone and a Mac with no server in the middle — over Bluetooth LE when that is all there is, over Wi-Fi when speed matters, switching between them without dropping the session.

Swift, Kotlin, Bluetooth LE, Wi-Fi TCP, AES-256-GCM

All projects

[04]

What I can do

No skill bars. Each of these is something I built, and the link goes to how.

Chained two detectors into one video pipeline: Vehicle detection feeds plate detection feeds OCR, with tracking in between so the expensive stage runs once per object instead of once per frame.Proof
Trained two detectors on 22,000+ images: Two YOLOv8n models on Indian road data — 20,800 IDD scenes and 1,618 plate images. The plate model reached 98.3% mAP@50.Proof
Built a RAG pipeline that is allowed to say no: Hybrid dense + BM25 retrieval fused by RRF, cross-encoder reranking, and a relevance floor that returns nothing rather than guessing — over eight policy PDFs, cited to the page.Proof
Shipped a full-stack app against a paid generative API: Django backend holding the Stability.ai key server-side, parsing the response and returning the image in one flow.Proof
Wrote a wire protocol across two native platforms: Swift on macOS and Kotlin on Android sharing one AES-256-GCM encrypted transport, with automatic handover between Bluetooth LE and Wi-Fi.Proof