AI/ML Developer · CSE Student · Hackathon Finalist

Priyanshu Yadav

Building ML systems, web apps, and real-world solutions.

Can you beat my reaction time?

Start — F1 test

PROJECTS

01

FinBud

AI · RAG · Finance 3rd — NIT Delhi, National

An AI-driven cross-border financial intelligence platform combining Retrieval-Augmented Generation with real-time streaming. Built to process multilingual financial documents from 14 countries with near-instant retrieval — eliminating the minutes-long indexing delay that made similar tools unusable in practice.

90% indexing time cut
<200ms document retrieval
1k+ files/week processed
14 countries supported
Python RAG Vector Search OCR Streaming Multilingual NLP
02

Nirbhaya

Geospatial · Safety · Full-stack

A community-powered safety intelligence platform that converts anonymous incident reports into real-time risk heatmaps. Instead of reacting to emergencies, it helps users navigate proactively — visualising danger zones by time of day before they travel. Built with a full geospatial pipeline: report ingestion, spatial aggregation, risk scoring, and automated data expiry to keep safety data relevant.

3 time-aware risk filters
SOS emergency interface
OSM open routing engine
Next.js TypeScript Supabase Leaflet OSRM Nominatim Tailwind
03

Pathfinder

ML · Route Generation

A machine learning system that generates optimal routes by learning from historical movement patterns, terrain data, and real-time constraints. Rather than relying purely on graph-based shortest-path algorithms, Pathfinder trains on past routing decisions to predict routes that reflect real human behaviour — accounting for factors that A* and Dijkstra ignore.

Python PyTorch Geospatial ML Graph Networks Route Prediction
04

F1 Reflex Game

Web · Frontend · Interaction

A Formula-1 inspired reaction tester that measures your response time to a race-start light sequence with millisecond precision. Built entirely in vanilla JS using the Performance Timing API — no frameworks, no libraries. Responsive across devices with instant feedback and scoring against real F1 driver benchmarks.

HTML CSS JavaScript Performance API

Recognition

Placed nationally.
Among 10,000+ builders.

Not just building for portfolio — competing, shipping, and ranking against thousands of engineers across India.

03 National rank, CodeSlayer 2025 NIT Delhi
10k+ Competitors in the hackathon
4 Live projects deployed

Code

From model to deployment.
No hand-offs.

Full-stack ML pipelines — training, serving, and monitoring. Everything from the data layer to the API, written and deployed by one person.

from fastapi import FastAPI, UploadFile from pipeline import embed_document, query_rag app = FastAPI() @app.post("/ingest") async def ingest(file: UploadFile): # OCR → chunk → embed → store chunks = await embed_document(file) return {"indexed": len(chunks)} @app.get("/ask") async def ask(q: str): return await query_rag(q)

Skills

The tools, not the buzzwords.
Things used in real projects.

Python C++ JavaScript TypeScript TensorFlow Keras PyTorch Pandas React Next.js Node.js Django MERN Firebase Tailwind SQL Git Linux
Bennett University, Delhi — B.Tech CSE(Specialization: Artificial Intelligence) · CGPA 8 · Graduating May 2028