VoiceGraph: Speak your automation, watch it execute. Built with Cerebras (Llama 4) for lightning-fast inference, Groq Whisper for voice input, and Docker MCP Gateway orchestrating secure containerized tools. Features visual debugging with AI error analysis, voice/text editing, and background scheduling.

Sure AI is an intelligent multi-agent platform that lets businesses deploy AI-powered customer support, email marketing, and recruiting — all in one place. At its heart is an embeddable chat widget that you can plug directly into your website, giving users instant access to your AI support agent (with integrations like stripe, slack and cal.com, let the agent manage it all for you). Each agent is powered by Cerebras and Meta models working in harmony — fast, context-aware, and deeply personalized. From resolving support tickets to writing persuasive emails to conducting interviews, Sure AI transforms how businesses communicate with people. It’s like giving your business a team of expert assistants that never sleep — one handling customers, one designing perfect marketing emails(multi-agent with research), and one interviewing candidates — all powered by blazing-fast Cerebras inference and the brilliance of Meta’s Llama models.

SRE Sentinel is an AI-powered monitoring and self-healing system for containerized applications. It continuously observes Docker containers, detects anomalies using advanced AI analysis, performs root cause analysis, and automatically applies fixes through the Model Context Protocol (MCP). The system provides real-time visibility into container health, resolves common issues automatically, and generates clear, human-friendly incident explanations for stakeholders.

SyncUp - AI Meeting Assistant for Google Meet A Chrome extension that captures conversations in Google Meet and automatically generates contextual information cards using AI. Built with Gemini API for transcription and Cerebras with Meta Llama for Fast and intelligent content generation. What It Does: SyncUp listens to your Google Meet conversations and automatically creates information cards about topics mentioned. When someone talks about Docker, Kubernetes, or any technical concept, the extension generates a detailed explanation card with key points, use cases, and resources - like having ChatGPT instantly explain things during your meeting. The extension also includes an AI chatbox where you can ask questions about the meeting or any topic, and get context-aware responses. Features: 1. Real-Time Topic Detection 2. AI-Powered Information Cards 3. AI Chatbox 4. Clean Interface Upcoming Features (1 month): 1. Meeting Memory for post-meeting analysis, questions, and cross-meeting reference 2. Multi-format attachments like PDFs, JPGs, MP4s, PPTs, etc. 3. Enhanced UI with new themes and trending skins for Google meetings. 4. Multi platform support for Zoom, Microsoft teams, etc.

Dungeon Escape is an immersive, retro-styled text adventure game that showcases the power of generative AI in creating dynamic, personalized gaming experiences. Built for the GenAI hackathon, this project demonstrates how large language models can transform traditional gaming through adaptive storytelling and infinite content generation. GenAI Innovation At the core of Dungeon Escape is our advanced integration with Cerebras LLM technology, which serves as a dynamic game master capable of generating unique narratives in real-time. Unlike scripted adventures, our AI creates personalized dungeon layouts, puzzles, and storylines that adapt to each player's choices, ensuring no two playthroughs are ever the same. Technical Implementation AI Architecture: Cerebras Llama 3.1 8B model for story generation Custom prompt engineering for consistent game mechanics Real-time response processing and validation Dynamic difficulty adjustment based on player performance GenAI Features: Context-aware narrative generation Adaptive puzzle creation Personalized story progression Multiple branching pathways and endings How GenAI Powers the Experience Dynamic Content Generation: The AI creates unique dungeon scenarios, characters, and challenges for each session Adaptive Storytelling: Narrative evolves based on player decisions and previous actions Intelligent Game Mastering: AI serves as dungeon master, responding to both predefined options and custom player inputs Progressive Complexity: Game difficulty scales with player demonstrated skill level Technical Stack AI Layer: Cerebras Cloud API, Custom prompt engineering, Real-time response parsing Backend: Node.js, Express, WebSocket communication Frontend: React, TypeScript, Retro-styled interface GenAI Challenges Solved Maintaining narrative consistency across AI-generated content Ensuring game balance while allowing for emergent storytelling Processing and validating AI responses in real-time Creating engaging puzzles that feel hand-crafted rather than generated Innovation in AI Gaming Dungeon Escape represents a new frontier in AI-assisted game development, demonstrating how generative models can create compelling, replayable experiences without extensive manual content creation. Our approach shows that AI can serve as both content generator and game master, opening possibilities for more dynamic and personalized gaming experiences across genres. This project highlights the potential of GenAI to transform not just what games contain, but how they're designed and experienced by players.

Mockly is an AI-powered mock interview platform that conducts realistic interviews and delivers intelligent, structured feedback. It uses Cerebras-hosted Llama models to generate adaptive, role-specific questions and a comprehensive post-interview assessment with category scores, strengths, and areas for improvement. Real-time voice is supported via Cartesia TTS and browser STT, and user sessions/results are persisted with Firebase. Built on Next.js (App Router) with a clean UI for fast practice and measurable improvement.

Project Title: Objective Email Problem Statement: Professionals and students receive dozens (sometimes hundreds) of emails every day. Important ones often get buried under newsletters, promotions, or automated notifications. Manually going through each email wastes time, causes missed deadlines, and reduces productivity. Existing email clients provide search and filtering, but they don’t actually help users cut through the noise to quickly identify what matters most. Solution – Objective Email: Objective Email is a productivity-focused email assistant that automatically fetches a user’s Gmail messages, stores them securely, and processes them intelligently. It directly addresses the problem of “information overload” in email by: Batch Fetch & Store: Our backend integrates with Gmail API to fetch and continuously update only new emails (avoiding duplicates), ensuring users always see the latest inbox snapshot. Centralized Storage: Emails are stored in a structured database with metadata like sender, subject, received time, tags, and status, making them queryable and filterable. Smart Summaries: Instead of forcing users to open long emails, Objective Email uses local ML models (no external API, protecting privacy) to generate quick summaries. Users can choose between modes like “Chill mode” (casual), “Professional tone,” or “Important-only.” Prioritization: Users can mark emails as important, and the system learns over time to highlight similar ones — reducing the chance of missing deadlines. Hackathon-Ready Reliability: To avoid Gmail’s timestamp issues, we designed a robust backend that deduplicates by messageId, ensuring refresh always captures new emails without errors. Real-Life Impact: Objective Email saves time and mental energy for students, professionals, and anyone drowning in emails. Instead of spending hours scanning irrelevant emails, users can open the app, hit refresh, and immediately focus on what matters: project updates, client mails, or academic instructions. By solving the daily frustration of missed important emails, Objective Email directly boosts productivity and reliability in personal and professional communication.

Aurora is a real-time AI field assistant that provides instant safety guidance and operational support to industrial workers through simple phone calls. When a worker calls Aurora, they speak naturally about their situation - whether it's a gas leak emergency, equipment malfunction, or routine procedure question - and receive immediate, contextual guidance in under 500 milliseconds. The system works through familiar phone infrastructure (no apps needed), making it accessible to every worker regardless of their technical expertise. Supervisors monitor all interactions through a live dashboard where they can see conversations in real-time, receive critical alerts, and take over calls when needed. Every interaction is automatically logged to Firebase for compliance and continuous improvement. We built Aurora because industrial workers face situations where immediate expert guidance can mean the difference between a near-miss and a tragedy. Traditional solutions are too slow (human operators take 30-60 seconds to respond) or too limited (static IVR systems can't handle dynamic situations). Aurora combines the speed of AI with the oversight of human supervisors to deliver the best of both worlds.

Patiently is an AI-powered platform that simplifies complex medical documents into clear, actionable insights for patients. Using Meta’s Llama models, it reads and translates lab results, imaging reports, and doctor’s notes into plain English—highlighting abnormal or critical values, explaining their significance, and suggesting questions to ask your healthcare provider. Key Features: Upload PDFs or images of medical reports (with OCR support) Automatic document classification (lab, imaging, notes, etc.) Plain-language translation of medical jargon Color-coded visual dashboard with normal, abnormal, and critical indicators Personalized “Questions for Your Doctor” suggestions Secure data handling (encrypted, privacy-first) Trend tracking for chronic patients Built With: React + Tailwind (frontend), FastAPI (backend), PostgreSQL (encrypted), Meta Llama 3.3 (AI engine), Tesseract OCR (document parsing), python (backend). Vision: To empower every patient to understand their health information without needing a medical degree—turning anxiety into clarity, and data into decisions.

Textsnap is a lightweight Chrome extension designed for academics and researchers. It reads the content of a webpage and allows you to quickly generate summaries or translations of selected text. Additionally, it integrates with your Obsidian vault via the Obsidian MCP server, enabling you to directly save notes to markdown files without switching windows.

CropGPT is a full‑stack AI advisor for farmers, delivering near sub‑second answers via Cerebras Llama 3.1‑8B. It orchestrates 6 MCP tools (prices, weather, soil, pests, research, mandi trends) and supports 10 languages with voice and cultural context. The React frontend and FastAPI backend use agentic reasoning, RAG, and persistent workflows for crop selection, irrigation, pest management, and harvest timing.

MediFitMate is a WhatsApp health & fitness companion bot designed to support users in managing their workouts, meals, hydration, yoga sessions, and medication. With AI-powered insights. MediFitMate helps users maintain a healthier lifestyle through personalized guidance and real-time assistance.