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. hashtag#DDSAIChallenge2026 hashtag#AIChallenge hashtag#MortgageAI hashtag#FintechInnovation hashtag

I’m proud to share the Product of my Agentic Predictive Maintenance Copilot for Industrial Assets. 🚀 What started as an engineering problem evolved into a working AI-powered maintenance system: 🔌 Live Industrial Telemetry — connected machine monitoring 🧠 ML-Based Prediction — machine health & failure risk 📡 Anomaly Detection — identify abnormal conditions 🤖 AI Diagnosis — evidence-based fault analysis 💬 Maintenance Copilot — actionable recommendations 🛠️ Work Orders — turning insights into maintenance action The core idea is simple: Predict before failure. Act before downtime. This project brings together Mechanical Engineering + Industrial AI + Automation into one practical solution. ⚙️🤖 Dr Ramanth Polu Sathyaram Sannasi Lori Figueiredo Jay Harish Jethva Ahmed Raoofuddin Sadiya Kauser Ahmad Burt Reynolds 🙏 A special thank you to the judges and mentors who took the time to review, evaluate and provide valuable feedback throughout the project. Especially grateful to Mohammad Arshad, Decoding Data Science, DDS Business Circle and everyone who contributed to this journey.

For the Building Agentic AI Application Challenge, I built Figuro — a hands-on AI learning app focused on helping people learn how to solve problems instead of simply giving them the answer. It includes: 🤖 Personalized AI Tutor 🎯 Weak-point detection ✍️ Practice & guided problem solving 🧠 Adaptive learning 📚 Quizzes and flashcards 🤖 Multiple AI agents The app is now deployed and live! 🎉 🔗 App: https://figuro-virid.vercel.app/ Really happy with how the project came together, and grateful to Decoding Data Science for the challenge and learning experience! #DDS #AgenticAI #AI #Figuro

Sharing my final project: Forsa AI, a voice-first interview practice platform. What it does: Candidate picks a role (any role, even a custom title) and optionally uploads a CV → has a real live voice interview with an AI interviewer (GPT-4o via Vapi, WebRTC) → gets an instant scorecard from Claude, cross-referenced against their CV, plus a downloadable PDF action plan with free courses matched to their weakest areas. A few engineering decisions I'm happy with: • Two-model architecture: GPT-4o live for speed, Claude post-call for structured, schema-guaranteed scoring • "Smart Mute" — muting mid-answer tells the agent to wait, not skip the question • A 30-case automated eval suite (npm run eval) that verifies the real prompt/logic wiring, not mocks — catches real regressions like a dead course link before submission, not after 🔗 Repo: github.com/muntherh/forsa-ai-agent 🎥 Demo video: [https://drive.google.com/file/d/1bDp7rxoqwgjNMUu3cMur2ARPlpWFgViy/view?usp=drivesdk] 🌐 Live: [https://forsa-ai-agent.vercel.app] Thanks to everyone here for the feedback along the way — happy to answer questions about the Vapi/Claude wiring if anyone's building something similar

I’m excited to share that I’ve reached the Day 7 checkpoint of the Building Agentic AI Challenge with my project: 🤖 Nevin Chatbot – AI Math Learning Assistant I’ve spent the last seven days working on building an AI-powered learning assistant that can help students learn math by providing clearer explanations, step-by-step worked solutions, answer verification, and followup interactive questions. Here’s what I focused on for checkpoint 7: 🔹 Fine tuning LLM prompts for relevancy and clarity 🔹 Validating normal questions, incomplete questions, ambiguous questions, followup questions 🔹 Contextual handling & conversation improvements 🔹 Improving prompt-injection/security/privacy defenses 🔹 Error handling/reliability 🔹 Validating public UX 🔹 Organizing my code/docs/reporting/progress github repo 🔹 Recording an explainer video of the project I learned a lot through this process. What really hit home during this whole experience, reinforcing something I already suspected, is that making AI applications that actually work in the real world is a whole lot more complex than just linking a user interface to an LLM. There’s Prompt Engineering, testing, security, reliability, UX, documentation, and iteration. So much more! Check out Nevin Chatbot - AI Math Learning Assistant GitHub: https://github.com/nevinsingh1313/nevin-chatbot-ai-math-learning-assistant Project Page: https://nevin-cs.github.io/nevin-chatbot-ai-math-learning-assistant/ Live Application: https://nevin-chatbot-ai-math-learning-assistant.ai.studio/ Completing Day 7 marks the end of this challenge, but definitely not the end of the project. I’m looking forward to continuing to improve the assistant, learning more about AI engineering, and building more useful AI applications. 7 days. 1 project. Loads of learning. 🚀