arslanahmed.me · LinkedIn · Fiverr · Email
I like building things that have to work beyond the demo.
Software Engineer working across mobile, backend, and AI. From training ML models and working with hardware to shipping my own Android app and building products for international clients, I enjoy taking problems from idea to something people can actually use.
Currently focused on AI, software architecture, and building better systems.
SpendWise · Live · Google Play
Built and shipped solo as a production Android app. SpendWise combines expense tracking with AI features, including a RAG-based financial advisor grounded in the user's own financial data and Gemini Vision for extracting transactions from receipts. I also built a native Kotlin home screen widget for quick expense entry and a Next.js API layer to keep AI credentials off the client.
React Native Kotlin TypeScript Expo Firebase Gemini Groq Next.js RAG Jest
Built for AG Leaders as an AI-powered crop disease detection and smart irrigation system. I trained a two-stage EfficientNetB3 pipeline that first identifies the crop (11 classes, 99.15% accuracy) and then detects diseases within the confirmed crop (76 classes, 94.86% accuracy, 0.94 macro F1). The system connects the ML pipeline to a React Native app and ESP32-based irrigation hardware.
React Native Node.js TensorFlow Flask EfficientNetB3 ESP32 HuggingFace
Built an AI-powered recruitment platform for managing candidates, screening interviews, and recruiter workflows. The system uses LangChain and Groq for automated screening and candidate scoring, PostgreSQL for application data, and Stripe for subscription-based features. I built the platform across the Next.js frontend and Django REST backend.
Next.js Django REST PostgreSQL LangChain Groq Stripe Supabase
Medical AI system built around RAG and multi-agent architecture to produce grounded responses from a dedicated medical knowledge base. The system retrieves relevant medical literature through vector search, routes tasks across specialized agents, and provides source citations with generated answers. I also evaluated the retrieval and response pipeline to measure how well the system grounds its answers in the available knowledge.
Python RAG Multi-Agent Pinecone LangChain Gemini NLP Evaluation
Languages JavaScript · TypeScript · Python · Kotlin · C# · C++
Mobile React Native · Expo · Android SDK
Frontend React · Next.js · Tailwind CSS
Backend Node.js · Express · Django REST · Flask
Databases MongoDB · PostgreSQL · Firebase · MySQL
AI/ML TensorFlow · PyTorch · Keras · LangChain · Scikit-learn · Pandas · NumPy
Tools Git · Linux · Postman · Android Studio