Hey! I'm Dhruval, a techie focused on real-world problems. From transit apps to AI, I engineer systems that scale.

devto

github

Dhruval Anandkar  • Peerlist

Built ARIA, a dual-mode accessibility app (React Native + FastAPI on Jetson Orin Nano) with on-device AI inference and SQLite—no external DBs. Implemented real-time ASL + emotion recognition using a multi-model fallback (Gemini, OpenAI, Claude, YOLOv8) with ElevenLabs TTS. Developed obstacle detection + turn-by-turn navigation via WebSockets and Google Maps APIs with low-latency + tunnel support (Cloudflare/ngrok).

ARIA

• Engineered a real-time, civic-grade transit platform utilizing a full-stack monorepo (Node.js, Express, MongoDB, React, React Native), enabling live GPS fleet tracking and instant dispatcher-to-driver communication via Socket.IO. • Developed a cross-platform mobile rider app (Expo/React Native) featuring Google Places API routing, dynamic fare calculation, QR-code boarding, and "APT Assist"—a custom, offline-capable intent-classification chatbot for automated user support. • Architected secure, role-based web portals for automated dispatching and fleet management, paired with a high-performance public marketing site featuring automated system theming and an interactive 3D hero environment built with React Three Fiber.

Ashland Public Transit (APT) | Real-Time Civi

Autonomous Model Governance: Architected a self-healing pipeline using LangGraph and XGBoost that automates drift detection (KS-Tests) and retraining, reducing manual intervention in the ML lifecycle by 100%. High-Performance Distributed System: Engineered a multi-cloud infrastructure (Aiven, Render, Vercel) supporting high-frequency telemetry ingestion with sub-second retrieval via B-Tree indexing and TimescaleDB. Resilient Fault-Tolerance: Designed a "Graceful Degradation" layer that ensures 99.9% uptime by automatically falling back to direct-to-database persistence during asynchronous broker (Kafka/Redpanda) outages. Real-time Observability: Integrated a Chaos Engineering suite with Discord Webhooks, enabling live alerting and manual overrides for autonomous SRE orchestration loops.

Vajra: Autonomous MLOps & Self-Healing SRE Pi

🔗 https://job-genie-ai-five.vercel.app Developed a full-stack SaaS application leveraging OpenAI GPT-4 and FastAPI to analyze resumes and recommend personalized job matches using embedding similarity and LLM agents. Engineered dynamic React frontend with TailwindCSS and real-time dashboards displaying AI-driven job fit scores, filters, and multi-user authentication via Firebase/Clerk.dev. Integrated advanced NLP pipelines including resume parsing, vector search with Pinecone, and LangChain orchestration to generate tailored role suggestions, skill gap analysis, and cover letter drafts. Deployed scalable backend on Render and frontend on Vercel, incorporating MongoDB Atlas for data storage and Stripe API for optional premium subscription monetization.

✨ JobGenie AI — Smart Resume Matching Web App

Developed a full-stack web app to analyze resumes using AI and match them with job descriptions. Integrated OpenAI's GPT API to generate match percentage and analysis feedback. Enabled secure user authentication using JWT and Express middleware. Implemented resume parsing with pdf-parse and mammoth for accurate content extraction. Deployed frontend on Vercel, backend on Render, and database on MongoDB Atlas. 🔐 Login with any username like testuser — password can be anything (demo only).

AI Resume Analyzer

Developed an AI-powered IoT system that collects real-time soil and sunlight data using sensors, enabling sustainable crop management through automated recommendations. Built and trained a machine learning model (TensorFlow, CNN) using Kaggle datasets; visualized performance with Matplotlib and evaluated results using accuracy metrics and confusion matrices. Designed a user-friendly interface for farmers and proposed future enhancements like plant disease detection and real-time analytics dashboard for smarter agriculture.

AgriScience – IoT-Embedded Solution for Agric

Built a full-stack web application using the MERN stack that enables users to auto-generate professional digital portfolios without any coding or design experience. Implemented a responsive, mobile-friendly interface using React.js and Tailwind CSS, allowing users to input personal/professional info and instantly share their portfolio via a public link. Integrated secure data handling and user customization features with real-time updates, enhancing usability and privacy.

FolioSpace – MERN Stack Digital Portfolio Gen

This weather forecast system allows users to search for real-time weather information by entering the name of a city. The system retrieves data such as the current temperature, humidity, wind speed, and weather conditions using the OpenWeatherMap API and displays it in a clean and user-friendly interface.

Weather Forecast Web Application

Optimized water usage with an automated irrigation system using Node MCU, ultrasonic sensors, and Arduino, enabling real-time monitoring for efficient irrigation via ThingSpeak data storage.

Water Management System (Node MCU, IoT)

Developed and maintained a comprehensive portfolio website showcasing technical projects, skills, and accomplishments, demonstrating proficiency in web development and design.

Portfolio Website (JavaScript, HTML/CSS) –

Developed IoT air quality system (Arduino, MQ-series sensors) to detect CO2 & PM2.5. Real-time data to cloud platform for informed actions.

Air Quality Monitoring (Arduino, IoT)

Autonomous Agents: Architected a multi-agent system using LangGraph and LangChain powered by Google Gemini to identify unexplored research trends and suggest experiment directions. Full-Stack Architecture: Built a high-performance backend using FastAPI (Python) and a responsive, type-safe frontend with React, TypeScript, and Tailwind CSS. Advanced Features: Implemented Google OAuth for secure authentication and integrated an automated export system to generate publication-ready .docx files. Data Management: Managed complex state using Zustand and TanStack Query, with a robust database layer powered by SQLAlchemy 2.0.

CortexLab: AI-Powered Research & Paper Genera

Redesigned the JoinNumi.com landing page as part of a frontend internship challenge with Bimi. The goal was to reflect the emotional and functional value of Numi AI — an AI-powered wellness coach — through a clean, responsive, and conversion-focused UI. Built using React, Tailwind CSS, Framer Motion, and Lottie animations, the site communicates Numi's mission to help users manage stress and burnout with small, personalized habits. 🔧 Tools: React, Tailwind CSS, Framer Motion, Lottie 🌿 Focus: Wellness-focused design, responsive layout, product storytelling 🔗 Live: https://bimi-numi-ai-redesign.vercel.app

Numi AI Landing Page Redesign

Built as part of a full-stack developer assessment in 3 days. Implemented a multi-step onboarding flow using: • Next.js + Tailwind CSS • Supabase (PostgreSQL) for database • Stripe for premium listing upgrade • Resend API for transactional emails • Google Places API for business search Includes Google business verification, user onboarding, payment integration, and email confirmation.

Swipe Savvy — Freemium Landing Flow

Created a Python-based automation tool that captures Google Form responses, summarizes them using OpenAI GPT-3.5, and categorizes feedback before posting to Slack via webhook. Integrated Google Sheets API, Slack API, and OpenAI API for seamless real-time feedback processing and internal communication. Enabled dynamic response categorization (e.g., Urgent, Follow-up) to streamline decision-making for product teams.

Google Form Feedback Summarizer – AI/ML Autom

Created a responsive front-end web application replicating Netflix India's user interface using HTML and CSS. Designed a visually appealing layout with categories, banners, and interactive elements to enhance user experience. Focused on clean, maintainable code and optimized styling for cross-device compatibility. Developed as a hands-on project to improve web design skills and UI/UX proficiency.