🥇 We Won Best Use of Snowflake at ELEMENTX 2026 ❄️ | MLH Hack Day 💻

A team of developers built ResolveIQ, an AI-powered incident copilot that won Best Use of Snowflake at ELEMENTX 2026, an MLH Hack Day hosted by Integral University's Department of CSE. The tool uses a RAG pipeline with Sentence Transformers and FAISS over seven PDF runbooks and ten historical incident tickets to retrieve context and generate cited root cause analyses, with Snowflake Cortex AI used for the hackathon implementation and Groq AI now handling LLM inference in the public repository. The team said the biggest lesson was that "building an AI application isn't just about choosing an LLM," citing context retrieval, grounding and sensitive-data handling as the bulk of the engineering work.

We won Best Use of Snowflake at ELEMENTX 2026 , an MLH Hack Day hosted by the Department of CSE, Integral University, Lucknow 🥳 Taking the top spot also came with a Machine Learning Kit 🛠️ and a .xyz domain 🌐 from XYZ. Definitely something we're super happy to receive 😊 In enterprise IT and FinOps, downtime can cost upwards of $9,000 per minute . When an incident happens, the information needed to understand it is often scattered across runbooks, documentation, historical tickets and system records. Engineers can spend significant time searching through all of this before figuring out what went wrong. We took on Problem Statement 02: "From Signal to Decision" to tackle this. ResolveIQ is an AI-powered incident copilot that brings relevant incident information together and helps engineers understand: It retrieves relevant context from runbooks and historical incidents, then generates a structured Root Cause Analysis RCA with supporting source references. For the Best Use of Snowflake category, we integrated Snowflake Cortex AI into our hackathon implementation to analyze incident context and generate cited RCA results. We also added a privacy layer for sensitive information and an incident command center for SLA tracking and remediation. The public repository has since evolved, and the current implementation uses Groq AI for LLM inference while retaining the core RAG and incident-resolution workflow. The core workflow is a RAG pipeline: 📄 Runbooks + Historical Tickets ↓ Document Processing ↓ Sentence Transformers ↓ FAISS Search ↓ Relevant Context ↓ AI Analysis ↓ 🧠 Root Cause Analysis ↓ ⚡ Next Steps The current project works with 7 PDF runbooks and 10 historical incident tickets, using all-MiniLM-L6-v2 embeddings and FAISS for semantic retrieval. We didn't want ResolveIQ to be just another chatbot. The interface includes: This lets engineers move from investigating an incident to understanding the evidence and deciding what to do next. AI: Snowflake Cortex AI, Groq Backend: FastAPI RAG: LangChain, FAISS, Sentence Transformers Frontend: React, Vite, Tailwind CSS The biggest takeaway for me was that building an AI application isn't just about choosing an LLM. A lot of the engineering work went into retrieving the right context, grounding responses in source material, handling sensitive data and turning AI output into something actually useful for engineers. This was definitely a team effort. Really grateful to have built this with Arshil Masood, Kamran Rizvi and Ayushmaan Vaibhav. 💙 We bounced ideas around, worked through problems, fixed what broke and kept pushing until we had something we could confidently put in front of the judges. 🥇 Best Use of Snowflake — ELEMENTX 2026 Huge thanks to Major League Hacking, Snowflake, XYZ and the Department of CSE, Integral University for making the experience possible. Definitely a hackathon we'll remember. 🚀