JEYADEV R

Portfolio 2026

Jeyadev R

AI/ML Engineer · Full-Stack Developer · Applied Research

Tap for sound
Jeyadev R
Scroll down

Hi, I'm

AI Engineer

Jeyadev

R

AI ArchitectFull Stack EngineerMERN Stack
1+Years Experience
31+Projects Built
AI + Full StackSpecialist

Aspiring AI Engineer building systems across machine learning, computer vision, and full-stack product, not just slides.

Available for freelance & full-time opportunities.

Available for work

Based in Chennai, India

Available Remote / Hybrid / On-site

Jeyadev R

Jeyadev

Who I Am

Python·TensorFlow·PyTorch·XGBoost·LightGBM·Scikit-learn·FastAPI·React·Next.js·Node.js·MongoDB·PostgreSQL·Docker·FAISS·Gemini API·Python·TensorFlow·PyTorch·XGBoost·LightGBM·Scikit-learn·FastAPI·React·Next.js·Node.js·MongoDB·PostgreSQL·Docker·FAISS·Gemini API·

B.Tech student in AI and Data Analytics who learns by building. My work spans forecasting and risk models (air quality, mortality, cardiovascular risk), computer vision (number plate recognition, scene classification), NLP and RAG (a drug-food interaction classifier, a self-hosted RAG cluster), and applied finance (a commodity forecasting ensemble, a prediction-market trading bot). I also work as a full-stack developer, shipping production MERN and Next.js apps end to end, from UI to auth to deployment, and I prefer shipping a working version over polishing a slide. Outside of code, I am a Student Council Member at my college and currently interning across two cleantech companies.

Projects

All projects
  • ScamShield: Explainable Scam SMS Detector for India

    AI / Security

    Explainable, obfuscation-robust scam SMS detector for Indian messages (KYC, UPI, courier, digital-arrest scams). Undoes tricks like '0TP' and 'v e r i f y', highlights the words behind each verdict, and checks links for look-alike bank domains. Of 8 benchmarked models, the deployed calibrated linear SVM reaches 0.97 F1, matching DistilBERT.

    Tech stack: Python, scikit-learn, PyTorch, DistilBERT, FastAPI, React

    ScamShield project detailsScamShield source
  • PHYS-TRUST Lite: Trust-Adaptive Physics-Informed Air Quality Sensing

    AI / IoT

    Air quality estimation on one low-cost Arduino node (MQ135 + DHT11). A Kalman filter fuses sensor physics, room mass-balance physics and an LSTM-ensemble prior, weighting each by a live trust score, with a 90% conformal uncertainty band and fault detection. Live React dashboard, local Gemma 3 explanations, 9% lower error than fixed-weight fusion.

    Tech stack: Arduino, PyTorch, FastAPI, React, WebSocket, Ollama

    PHYS-TRUST Lite project detailsPHYS-TRUST Lite source
  • Attendance ERP: Multi-Semester College Attendance System on the Edge

    Full-Stack / EdTech

    College attendance system built around real timetables: multi-hour labs marked once, partial exam days, one-off period swaps, and bunk detection. Mobile-first marking, timetable builder with conflict checks, a student planner that projects when attendance crosses the limit, and admin heatmaps with an audit log, all on Cloudflare Workers + D1.

    Tech stack: Next.js, TypeScript, Hono, Cloudflare Workers, D1, Prisma

    Attendance ERP project detailsAttendance ERP source
  • FRONTIER: Local AI Content Agent for a Discord Community

    AI / Automation

    AI agent for a student tech community Discord server. Discovers hackathons, research and AI news, drafts posts in the author's writing style with a local LLM (Ollama / LM Studio), routes them to the right channel, and publishes only after human approval via a React dashboard or Discord buttons. Ships an MCP server for other AI agents.

    Tech stack: Python, FastAPI, React, Ollama, discord.py, MCP

    FRONTIER project detailsFRONTIER source
  • ASPCV CRM: Full-Stack CRM for a Cleantech Company

    Full-Stack / SaaS

    Full-stack CRM covering the full customer lifecycle from lead to deal, project, installation, and support, plus HR/payroll/attendance, inventory, material requests, invoicing, and dynamic database-backed role-based access control.

    Tech stack: React, TypeScript, Node.js, Express, Prisma, PostgreSQL

    ASPCV CRM project detailsASPCV CRM source
  • sret-rag: Self-Hosted RAG on a 4-PC LAN Cluster

    AI / Infrastructure

    Built for offline AI access at college with zero cloud dependency. Fully offline, self-hosted Retrieval-Augmented Generation system distributed across 4 budget PCs on a LAN: a FastAPI gateway, dedicated embedder, FAISS vector index, and a llama.cpp/Qwen2.5 LLM node. ChatGPT-style UI, cited answers, embed caching, doc dedup, and session-based auth.

    Tech stack: FastAPI, FAISS, llama.cpp, Qwen2.5, SQLite, sentence-transformers

    sret-rag project detailssret-rag source
  • PolyBot: Autonomous Prediction-Market Trading Engine

    AI / Trading

    Autonomous trading engine for Polymarket prediction markets, with live order book/price ingestion, mispricing/momentum/news strategy scoring, on-chain execution via Web3, a risk management core, and a control-panel dashboard for monitoring and steering the bot.

    Tech stack: Python, FastAPI, Web3, Redis, PostgreSQL, Docker

    PolyBot project detailsPolyBot source
  • AbsorbAI: Clinical Pharmacokinetic Decision-Support Platform

    AI / Healthcare

    Clinical decision-support platform simulating personalized drug behavior across ~45 drugs for doctor and patient roles. Features a 10-layer PK/PD engine, Monte Carlo uncertainty quantification, Bayesian updates, a pharmacological knowledge graph, and Gemma 1.5B LLM integration.

    Tech stack: Python, Django, NetworkX, Gemma 1.5B

    AbsorbAI project detailsAbsorbAI source
  • Pollumap V3: Real-Time Air Quality Dashboard

    Full-Stack / Environmental

    Full-stack air quality dashboard for Chennai with live AQI charts, an interactive Leaflet map, XGBoost-powered 24-hour forecasts, an A* 'Healthy Routing' algorithm for least-polluted paths, source-apportionment breakdowns, and an IoT sensor dashboard.

    Tech stack: React, Vite, Flask, MongoDB, XGBoost, Leaflet

    Pollumap V3 project detailsPollumap V3 source
  • Market Oracle: Commodity Price Forecasting Ensemble

    AI / Finance

    Commodity price prediction pipeline for the Mitsui Commodity Prediction Challenge, combining a 9-model ensemble (Chronos, NeuralProphet, NeuralCDE, Mixture-of-Experts, Temporal Graph Attention, Granger causality, regime detection) via a Bayesian meta-ensemble with confidence intervals.

    Tech stack: PyTorch, NeuralProphet, Chronos, Transformers, NetworkX

    Market Oracle project detailsMarket Oracle source
  • EcoSynth: Neuro-Symbolic Green Retrosynthesis Planner

    AI / Chemistry

    AI-assisted retrosynthesis planner combining symbolic chemistry rules, RAG over literature/reaction data, and a green-chemistry scorer to propose, validate, and rank synthesis routes, with built-in hallucination detection and a React UI for exploring reasoning traces.

    Tech stack: Python, FastAPI, ChromaDB, React, ChemBERTa

    EcoSynth project detailsEcoSynth source
  • HyperlocalAQ: Multi-Source Air Quality Modeling for Chennai

    AI / Environmental

    High-resolution air quality modeling framework integrating 7 APIs (WAQI, NASA FIRMS, Open-Meteo, TomTom, OSM, CAMS, OpenFDA) with hybrid IDW modeling, fire-impact and traffic adjustments, visualized on an interactive Folium map.

    Tech stack: Python, Folium, IDW Modeling

    HyperlocalAQ project detailsHyperlocalAQ source
  • Aspiration Cleantech Ventures: MERN Web Platform for a Solar Energy Company

    Full-Stack / SaaS

    Production MERN-stack web app built during a full-stack internship for a cleantech/solar energy company. JWT authentication, Helmet security hardening, and Cloudinary-based image uploads.

    Tech stack: React, Vite, TypeScript, Node.js, Express, MongoDB

    Aspiration Cleantech Ventures project detailsAspiration Cleantech Ventures source
  • sret-noticeboard: College Notice Board with Admin Panel & Calendar

    Full-Stack / EdTech

    Built for my college's notice board system. Role-based auth (student/faculty/admin), an admin dashboard with usage stats, image-attached notices, and a FullCalendar-driven event calendar. Node/Express/MongoDB backend with JWT auth and Multer-based image uploads.

    Tech stack: Node.js, Express, MongoDB, JWT, Multer, FullCalendar

    sret-noticeboard project detailssret-noticeboard source
  • Cardio Disease Detection Engine: Cardiovascular Risk Dashboard + API

    AI / Healthcare

    ML system predicting cardiovascular disease risk (High/Low) from patient clinical and lifestyle data. Streamlit dashboard and Flask API backed by a tuned model zoo (Logistic Regression through stacking/voting ensembles), with the best model auto-selected for production.

    Tech stack: Python, Streamlit, Flask, XGBoost, LightGBM, CatBoost

    Cardio Disease Detection Engine project detailsCardio Disease Detection Engine source
  • AeroSight: Air Quality Forecasting & Health Reporting Toolkit

    AI / Environmental

    Air quality analysis and forecasting toolkit combining Prophet and XGBoost dual-model forecasting, correlating pollution with mortality data, and auto-generating multi-page PDF reports for monitoring stations.

    Tech stack: Python, Prophet, XGBoost, Open-Meteo

    AeroSight project detailsAeroSight source
  • AI Drug-Food Interactions: NLP-Powered Interaction Classifier

    AI / Healthcare

    NLP + ML web app classifying drug-food interactions into 6 categories using TF-IDF vectorization and ensemble models (99.2% accuracy), with semantic search and a generated knowledge graph.

    Tech stack: Python, Streamlit, Scikit-learn, TF-IDF

    AI Drug-Food Interactions project detailsAI Drug-Food Interactions source
  • Intel Scene Classification: CNN vs VGG16 vs Bagged Ensemble

    AI / Computer Vision

    Image classification on the Intel Image Classification dataset (6 scene classes), comparing a baseline CNN, fine-tuned VGG16 transfer learning, and a bagged-ensemble of Dense heads (~93% test accuracy), with Grad-CAM visualizations and misclassification analysis.

    Tech stack: TensorFlow, Keras, scikit-learn, Grad-CAM

    Intel Scene Classification project detailsIntel Scene Classification source
  • GloVe from Scratch: Word Embeddings Trained on WikiText-2

    AI / NLP

    GloVe word embeddings built from first principles: sparse co-occurrence matrix (120x smaller than dense), hand-derived gradients verified against autograd, and manual AdaGrad, with no autograd or torch.optim. Trained on WikiText-2 in 26 s on a 4 GB laptop GPU, with analogy tests, ablations and a comparison against official GloVe vectors.

    Tech stack: Python, PyTorch, CUDA, NumPy

    GloVe from Scratch project detailsGloVe from Scratch source
  • ANPR-YOLOv4: Automatic Number Plate Recognition

    AI / Computer Vision

    Automatic Number Plate Recognition system using a custom-trained YOLOv4 detector for cars, bikes, and autos. Custom-labeled dataset, Darknet training pipeline, and inference notebook for detecting vehicles and localizing their number plates.

    Tech stack: YOLOv4, Darknet, OpenCV, Python

    ANPR-YOLOv4 project detailsANPR-YOLOv4 source
  • Cardiovascular Risk Prediction: ML Pipeline for Patient Health Data

    AI / Healthcare

    End-to-end ML pipeline predicting cardiovascular disease risk from patient health data. Engineered features (BMI, MAP, pulse pressure, hypertension flag, heuristic risk score) feed baseline and tree-based models, evaluated on accuracy, precision, recall, F1, and ROC-AUC.

    Tech stack: Python, scikit-learn, XGBoost, LightGBM, CatBoost, TensorFlow

    Cardiovascular Risk Prediction project detailsCardiovascular Risk Prediction source
  • T20 Fantasy Impact Model: LightGBM Fantasy Team Optimizer

    AI / Sports Analytics

    ML-driven T20 cricket fantasy team optimizer. Dual LightGBM models (mean + P90 quantile) trained on ball-by-ball data predict batting runs and bowling impact; a match-day engine then builds an optimal Dream11-style XI under credit/role constraints with captain/vice-captain selection.

    Tech stack: Python, LightGBM, pandas, joblib

    T20 Fantasy Impact Model project detailsT20 Fantasy Impact Model source
  • Aspiration Cleantech Academy: Job-Oriented Training Platform

    Full-Stack / EdTech

    Job-oriented training platform covering HVAC, entrepreneurship, digital marketing, and sales courses. Next.js frontend with an Express/MongoDB backend, built during a digital marketing internship landing-page build.

    Tech stack: Next.js, TypeScript, Express, MongoDB

    Aspiration Cleantech Academy project detailsAspiration Cleantech Academy source
  • VoteIQ: AI-Powered Civic Education App

    AI / Civic Tech

    AI-powered civic education assistant helping first-time Indian voters understand the election process. Built in 8 hours at Google's Prompt Wars, featuring a guided voter journey, myth-busting AI module, guarded Q&A assistant, civic maturity scoring, and a readiness quiz.

    Tech stack: Flask, Vanilla JS, Gemini 2.0 Flash, Google OAuth

    VoteIQ project detailsVoteIQ source
  • Advanced CLI Web Scrapper: Full-Site Archiver Built on Playwright

    Tooling

    CLI tool to download full web pages with assets (CSS, JS, images, video) for offline archiving. Handles dynamic/JS-heavy pages, adjustable crawl depth, URL filtering, domain locking, and concurrent downloads.

    Tech stack: Python, Playwright, BeautifulSoup

    Advanced CLI Web Scrapper project detailsAdvanced CLI Web Scrapper source
  • InstaAnalyzer: Follower/Following Analysis Script

    Tooling

    Python script analyzing Instagram accounts to identify mutual followers and non-followbacks, using Instaloader for data and the Rich library for clean terminal output.

    Tech stack: Python, Instaloader, Rich

    InstaAnalyzer project detailsInstaAnalyzer source
  • Quantum Teleportation Simulation: Interactive Browser-Based Quantum Circuit Visualizer

    Education / Visualization

    Zero-dependency browser simulator visualizing the quantum teleportation protocol step-by-step, with selectable noise models, an animated quantum circuit, and real-time fidelity metrics.

    Tech stack: JavaScript, Tailwind CSS, SVG

    Quantum Teleportation Simulation project detailsQuantum Teleportation Simulation source
  • Sudoku Visualisation: Algorithm Comparison: Backtracking vs MRV

    Algorithms / Visualization

    Graphical Sudoku solver and visualizer built natively in C++ with SFML. Generates puzzles at three difficulty levels and runs a side-by-side timing comparison of Backtracking vs MRV-heuristic solving.

    Tech stack: C++, SFML

    Sudoku Visualisation project detailsSudoku Visualisation source
  • Arduino Calc: Matrix Edition: Von Neumann CPU Concepts on Real Hardware

    Embedded / COA

    Calculator built for a Computer Organization & Architecture course, mapping Von Neumann CPU concepts onto Arduino Uno hardware with a matrix-styled Web Serial interface that needs no Serial Monitor.

    Tech stack: Arduino, Web Serial API, HTML/CSS

    Arduino Calc: Matrix Edition project detailsArduino Calc: Matrix Edition source
  • Urban Farming Simulator: OOP Rooftop Farm Management Sim

    Academic / Java

    Text-based rooftop urban farm management simulation modeling crop biology, seasonal weather, and market fluctuations using SOLID principles, encapsulation, and polymorphism. Built with a 3-person team.

    Tech stack: Java, OOP

    Urban Farming Simulator project detailsUrban Farming Simulator source
  • Crorepati Quiz Game: CLI Quiz Game Inspired by KBC

    Academic / Java

    CLI-based quiz game inspired by 'Kaun Banega Crorepati', built in Java with lifelines and escalating prize levels.

    Tech stack: Java

    Crorepati Quiz Game project detailsCrorepati Quiz Game source
  • StyleGAN Exploration: Style-Based GAN in TensorFlow/Keras on CIFAR-10

    AI / Deep Learning

    Compact StyleGAN implementation in TensorFlow/Keras for image generation on CIFAR-10, in a single Jupyter notebook. An 8-layer mapping network and AdaIN style modulation drive synthesis, with style mixing and latent interpolation experiments. An educational implementation, not a reproduction of StyleGAN2 or StyleGAN3.

    Tech stack: Python, TensorFlow, Keras, Jupyter

    StyleGAN Exploration project detailsStyleGAN Exploration source
  • Design and Analysis of Algorithms Study Guide: Student guide to 16 classic algorithms with C code

    Education / Algorithms

    PDF study guide and C implementations for 16 key algorithms and problems from Anany Levitin's Introduction to the Design and Analysis of Algorithms: searching, sorting, N-Queens, knapsack, Kruskal, Dijkstra, Graham scan, Miller-Rabin and LCS, organised into Basic, Moderate and High difficulty levels.

    Tech stack: C, Algorithms, Data Structures

    Design and Analysis of Algorithms Study Guide project detailsDesign and Analysis of Algorithms Study Guide source
  • ASPCV WordPress Site: Custom Astra child theme and mu-plugins for a cleantech company site

    Full-Stack / WordPress

    Source-controlled custom code for the corporate website of Aspiration Cleantech Ventures: an Astra child theme with custom page templates, plus must-use plugins for the navigation system and redirects, deployed on WordPress.

    Tech stack: WordPress, PHP, Astra, Elementor, LiteSpeed

    ASPCV WordPress Site project detailsASPCV WordPress Site source
Work Experience05 Companies
01
May 2025 - Jul 2025InternshipChennai, India
Sri Ramachandra Faculty of Engineering and Technology

Sri Ramachandra Faculty of Engineering and Technology

Machine Learning Intern

  • Built AeroSight, a Prophet + XGBoost dual-model air quality and mortality forecasting toolkit with auto-generated PDF reports, placing 2nd at Research Day 2025.
PythonXGBoostProphetScikit-learnData Analysis
02
Jan 2026 - Mar 2026InternshipChennai, India
Aspiration Cleantech Ventures

Aspiration Cleantech Ventures

Full-Stack Developer (Intern)

  • Shipped a production MERN web app (React/TypeScript + Node/Express/MongoDB) with JWT auth and Cloudinary uploads.
ReactTypeScriptViteNode.jsExpressMongoDBJWT
03
May 2026 - PresentInternshipChennai, India
Aspiration Cleantech Ventures

Aspiration Cleantech Ventures

SAAS Developer (Intern)

  • Building SaaS features and REST APIs for new product modules on the core MERN platform.
ReactNode.jsMongoDBREST APIs
04
May 2026 - PresentInternshipChennai, India
Aspiration Cleantech Academy

Aspiration Cleantech Academy

Digital Marketing Intern

  • Built and deployed the academy's training website (Next.js + Express/MongoDB) on a purchased domain via Cloudflare; promoted webinars and HVAC/cleantech course content across Instagram, LinkedIn, Facebook, and YouTube.
Next.jsCloudflareContent StrategyDigital Marketing
05
Jul 2026 - PresentLeadershipChennai, India
Sri Ramachandra Institute of Higher Education and Research

Sri Ramachandra Institute of Higher Education and Research

Student Council Member

  • Representing students on the council, coordinating with faculty on campus initiatives and student affairs.
LeadershipEvent Coordination
AI ENGINEER

JEYADEV R

Jeyadev R is an AI engineer and full-stack developer in Chennai, India, building machine learning, computer vision, RAG and production web systems.

Good afternoon

AI Engineer

Jeyadev R

Jeyadev R is an AI engineer and full-stack developer in Chennai, India, building machine learning, computer vision, RAG and production web systems.

jeyadev2006@gmail.com

Available for collaborations

Crafting modern
digital products
that feel
alive.

Let's talk →

JEYADEV R

JR

© 2026 JEYADEV R

ALL RIGHTS RESERVED

DESIGNED & DEVELOPED
WITH PRECISION.
✺