DDS Data Challenge Project Showcase
19 AI applications, built and deployed by our community during the 8-day DDS Data Challenge (Sept 5–13) — from idea to live demo, GitHub repo, and walkthrough video. Explore every build below.
GenNPost
Multi-agent system that automates content creation and publishing across multiple platforms.
Nahwi Fix
An Arabic grammar-correction app that helps users write with correct grammar in real time.
MortgageAI
Automated eligibility & underwriting engine — calculates FOIR/EMI loan eligibility and generates personalized underwriting checklists using n8n + Claude + Google Sheets/Drive.
QualityFlow Lite
Agentic web app that converts a short manufacturing defect description into a structured preliminary NCR/CAPA draft.
GuardianDrive AI
Real-time driver safety system using computer vision and Eye Aspect Ratio tracking to detect fatigue and trigger instant alerts.
GroupMate
AI-powered task distribution app for student group projects — splits tasks fairly by skill fit and workload, tracks progress, flags anyone falling behind.
Abstratly AI
Lets a researcher input their field or paper abstract, then sends weekly updates on newly published related papers.
BuildOS
Agentic AI project management platform that turns ideas and goals into structured, actionable project plans and priorities, guiding users through execution.
Evidence Investigator AI
Investigates suspicious URLs, emails, text, and files by gathering cybersecurity evidence and combining threat-intelligence data with AI reasoning to produce explainable risk scores.
ProofCheck
AI evidence-reconciliation agent — investigates whether a business claim is supported by evidence across multiple documents using RAG, agentic investigation, and deterministic validation.
Navigate
Analyzes links and surfaces the requirements and details behind them.
AgroVn AI — Smart Agronomist Assistant
Gives farmers safe, actionable AI guidance for their crops and operations.
CareLoop
Agentic AI companion that calls elderly users to check on medications, listening for real signs of trouble, and reasons across days of check-ins, drug interactions, diet and vitals to decide when to alert family — with a caretaker dashboard.
Figuro
Personalized AI learning coach that helps students understand, practice, and solve problems using their own study materials, tracking weak points and adapting practice.
Forsa AI
Autonomous voice-first AI mock interviewer conducting dynamic spoken job interviews in English, adapting questions in real time and delivering instant performance scorecards.
ProSight AI V2
Second version of a construction project intelligence platform, upgraded and enhanced.
ProDiag AI V2
Agentic predictive-maintenance copilot for industrial assets.
Kindred
No-code AI-enabled community assistance platform connecting seniors, people with disabilities, and people living alone with trusted helpers for errands, transportation, medical appointments, and companionship.
Nevin Chatbot — AI Math Learning Assistant
AI-powered educational assistant that helps students understand and solve math problems through interactive, step-by-step learning instead of just giving final answers.
No projects match your filter.
Hosted by Mohammad Arshad
Founder of Decoding Data Science and host of the DDS Data Challenge. Every project here was created by a participant who moved from idea to a working, deployed AI application — the DDS way: learning by building. With 194,000+ community members across 50+ countries, DDS helps builders ship real AI products.
Want to ship an AI app like these?
Join the DDS community and our upcoming programs to go from idea to deployed AI application — with mentorship, a CPD-backed certificate, and a room full of builders.
Explore DDS Programs →Projects shown are participant submissions from the DDS Data Challenge (Sept 5–13). Some demo, video, or repository links may be added or updated over time. Certificates and any rewards are subject to eligibility, project quality, submission completion, and final judge review.
🚀 From an idea… to an application with a purpose. Meet Kindred. ❤️
As the DDS September 2026 Agentic AI Challenge comes to a close, I’m incredibly grateful for the opportunity to build Kindred — an AI-powered concept built around something technology should never lose sight of: human connection. Kindred may have started as an 8-day challenge project. But the problem it hopes to solve exists every single day.
Kindred was created with a simple vision: help seniors, people living alone, individuals with disabilities, or anyone who needs a helping hand connect with people who can assist with everyday needs — from errands and appointments to transportation and companionship — through paid or volunteer support.
But this challenge became about much more than building an application. It was about learning, experimenting, failing, fixing, rebuilding, sharing, and discovering what becomes possible when AI is applied to a real human problem.
A huge thank you to Decoding Data Science (DDS), Mr. Mohammad Arshad, the mentors, judges, and this incredible community for creating an environment where people aren’t just learning about AI — they are building with it.
And congratulations to every participant. Regardless of rankings, everyone who had the courage to turn an idea into something real is already a winner.
Congratulations dear Ajoy !!
MortgageAI — Automated Mortgage Underwriting. Accurate. Deterministic. Anytime.
Thrilled to share my final submission for the DDS AI Challenge 2026 — MortgageAI, a fully automated, no‑code/low‑code mortgage underwriting engine built end‑to‑end using n8n, Anthropic Claude Sonnet, OpenAI GPT5/GPT4o, and Google Cloud.
MortgageAI delivers FOIR, EMI, loan eligibility, multi‑bank underwriting, document checklists, and email-ready summaries —
What I Built
Deterministic FOIR/EMI engine
Multi‑bank underwriting logic (HDFC, ICICI, SBI; Axis/Kotak next)
AI‑powered PDF parsing with fallback models
Automated Drive folder creation
Google Sheets logging
Email delivery with underwriting summary
Fully orchestrated through n8n
Multi‑Model Strategy
Claude Sonnet 4.5 → Underwriting logic
GPT5 / GPT4o → PDF parsing fallback
Deterministic JSON schema → Zero hallucinations
Performance
Underwriting latency: 1.8–2.4 sec
Workflow latency: 3.5–5 sec
End‑to‑end: ~6–7 sec
Cost/run: $0.002–$0.006
Live Demo & Links
n8n Workflow Demo: https://lnkd.in/gHwbpTBs
Project Website: https://lnkd.in/dVXxMrUb
Repo: https://lnkd.in/dfKxfgaP
This challenge pushed me to design a production‑grade underwriting engine using a no‑code/low‑code stack — and I’m proud of the final outcome.
Grateful to Decoding Data Science, the mentors, and the entire community for the energy throughout the challenge.
Great.. congrats
It was an inspiring and wonderful journey building and developing the ‘Forsa’ platform during this exceptional challenge. Huge thanks and appreciation to the organizers for the outstanding arrangement and for providing this valuable opportunity to learn and apply AI agent technologies. We always look forward to more challenges and initiatives that drive innovation forward. Wishing everyone the 🤞.
Congrats
Participating in the DDS AI Application Challenge was a valuable experience that pushed me beyond learning AI concepts into actually building and deploying a real-world application. Throughout the challenge, I learned how to turn an industrial problem into an end-to-end AI solution, while improving my skills in machine learning, LLM integration, system architecture, deployment, and product development. The guidance and community feedback helped me continuously refine my approach and think more practically about building AI applications.
For this challenge, I developed **ProDiag AI V2 — an AI-Powered Predictive Maintenance Copilot for Industry**. The project focuses on helping maintenance teams move from reactive maintenance to proactive decision-making by connecting **machine monitoring, anomaly detection, failure prediction, AI-powered diagnosis, and maintenance recommendations** into one workflow. I built the system with a focus on real industrial scenarios rather than treating it as only a demonstration or chatbot.
One of the most valuable parts of this project was taking ProDiag from an initial predictive-maintenance concept toward a more complete, production-oriented architecture. The project combines machine-learning models, real-time industrial telemetry, AI reasoning, and an interactive application to demonstrate how AI can support maintenance engineers and reduce unplanned equipment downtime.
Overall, this challenge gave me the opportunity to build, test, receive feedback, iterate, and think like an AI product builder. I’m grateful to **Decoding Data Science and Mohammad Arshad** for creating an environment where participants can turn ideas into working AI applications and continue improving them beyond the challenge.
MPTU, Nipun.. Great work as always.
Congratulations for the excellent project !!
Really enjoyed this challenge and the experience of building Figuro. It gave me the opportunity to explore agentic AI and think more deeply about how AI can support personalized learning. I learned a lot throughout the process, especially from turning an idea into a working application. Thank you Decoding Data Science for the opportunity and for creating such a great learning experience!
GroupMate: AI-powered task distribution for student group projects
Every student has lived this: group work gets split unevenly, and nobody notices a teammate falling behind until the deadline is already a crisis. GroupMate fixes that end-to-end, not just the assignment part.
What makes it different from a normal to-do/project tool:
– Tags built for the actual project. GPT-4o reads the real project description and generates skill tags specific to it. Not generic buckets like “teamwork” or “communication.”
– Automatic falling-behind detection. Compares each member’s real completion percentage against real time-elapsed, not a fixed check-in schedule. flags anyone trailing by more than 20%, visible to the whole team, before it’s a crisis.
– An AI assistant grounded in the team’s own files. RAG over uploaded documents, with the source cited. Not a generic chatbot bolted on for show.
On the engineering side: AI-reported hour estimates are clamped so a single hallucinated number can’t silently corrupt the fairness math, and the assistant is fenced against prompt injection, tested with a real injection attempt against my own file-search pipeline, not just assumed safe. I ran a 12-sample evaluation across wildly different project types (a documentary film, a Discord bot, a podcast series) and got valid, project-specific tag sets on all 12 before trusting it with real data.
Built solo across 8 days with Gradio, Supabase, and the OpenAI API, deployed live on Render.
🔗 Live demo: https://groupmate-glp1.onrender.com
💻 GitHub: https://github.com/ranasalahe/GroupMate
Thank you to Decoding Data Science and Mohammad Arshad for a challenge that forced actual shipping, not just planning.