ATMECS Global GEN AI Hackathon

live
4900 Registered Allowed team size: 2 - 5
4900 Registered Allowed team size: 2 - 5

The idea submission phase is over and participation is closed.

idea phase
Online
starts on:
Sep 02, 2024, 12:30 PM UTC (UTC)
ends on:
Sep 29, 2024, 06:29 PM UTC (UTC)
Prototype Phase
Online
starts on:
Oct 04, 2024, 12:30 PM UTC (UTC)
ends on:
Oct 16, 2024, 06:29 PM UTC (UTC)

Overview

Hacking is building things that you always wanted to have but no one has built it yet. It's to come up with an amazing idea and work tirelessly on it. It is to fail, fail again and fail better. Try out new things and learn while doing that. It's to work together, collaborate and build things that are innovative. It is to be a better programmer.

With that spirit, HackerEarth is conducting a hackathon - ATMECS Hackathon 2024 and invites all developers and hackathon enthusiasts to participate in an interesting and engaging hackathon.

Themes

Enterprise RAG - Transform Business Process & Decision-Making with Generative AI

Problem Statement Name: Revolutionizing Information Access and Decision-Making with Large Language Models and Retrieval-Augmented Generation

Overview: Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG) are set to redefine how enterprises make decisions. Your challenge is to develop a cutting-edge solution that leverages these Generative AI technologies to transform decision-making processes. Move beyond traditional systems and explore how LLMs can provide dynamic, context-aware insights and recommendations that empower stakeholders with innovative strategies and actions. Your solution should integrate RAG to enhance the LLM's capabilities with up-to-date, domain-specific knowledge, and incorporate a robust Q&A system for interactive decision support.

Tech Stack:

  • Python, JavaScript (React/Node.js)
  • Hugging Face Transformers, LangChain, OpenAI API
  • Large Language Models (e.g., GPT-4, BERT, T5)
  • Vector Databases (e.g., Pinecone, Weaviate, Faiss)
  • Cloud Platforms (AWS, Azure, Google Cloud)
  • Data Visualization Tools (D3.js, Plotly, Streamlit)

Restrictions:

  • Solutions must incorporate at least one Large Language Model and implement RAG.
  • Ensure data privacy and security measures are implemented, especially when handling sensitive business information.

Exact Task:

  • Mandatory Tasks:
    1. Develop a Gen AI-powered engine using an LLM that can analyze business data and suggest actionable strategies.
    2. Implement RAG to enhance the LLM's knowledge with company-specific data and industry trends.
    3. Build a user-friendly interface with real-time data visualization that reflects AI-driven insights.
    4. Create an interactive Q&A system that allows users to ask specific questions about business decisions and receive context-aware responses.
  • Good-to-Have Tasks:
    1. Implement few-shot learning to adapt the model to different business scenarios quickly.
    2. Develop metrics to evaluate the quality and relevance of AI-generated insights and Q&A responses (e.g., relevance scores, user feedback integration).
    3. Implement a system to track and measure the impact of AI-driven decisions on key business metrics over time.
  • Bonus Tasks:
    1. Develop a case study showing how your Gen AI solution impacts decision-making in a specific industry, including quantitative improvements in decision outcomes.

Submission Format:

  • Detailed workflow documentation.
  • Source code on GitHub/GitLab.
  • A demonstration video showcasing the solution in action.
  • PPT with the approach, results, and potential impact.

Resources:

  • Introduction to Large Language Models
  • Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
  • Building Effective Q&A Systems with LLMs
NIM (NVIDIA Inference Microservices) - Accelerate Gen AI Applications

Problem Statement Name: Scaling Generative AI with NIM

Overview: Leverage the power of NIM (NVIDIA Inference Microservices) to scale and optimize Generative AI applications. Your task is to create high-performance Gen AI solutions by integrating NIM services into cutting-edge applications or developing new tools that empower engineers to seamlessly incorporate NIM into their Gen AI workflows. Your solution should demonstrate how these microservices can be used to enhance AI functionalities, focusing on efficient deployment and scaling of large language models and other generative AI systems.

Tech Stack:

  • NIM (NVIDIA Inference Microservices)
  • Python, Go
  • Docker, Kubernetes, Helm
  • PyTorch, ONNX Runtime, TensorRT
  • Transformer models (e.g., BERT, GPT)
  • Microservices Architecture
  • Cloud Platforms (AWS, Azure, Google Cloud)
  • Monitoring tools (Prometheus, Grafana)

Restrictions:

  • Utilize at least two NIM microservices in your solution.
  • Focus on optimizing inference for large generative models.

Exact Task:

  • Mandatory Tasks:
    1. Integrate NIM microservices to deploy and scale a large language model or generative AI system.
    2. Design a tool that simplifies the deployment and optimization of Gen AI models using NIM.
    3. Implement dynamic batching and model parallelism for efficient inference.
  • Good-to-Have Tasks:
    1. Provide comprehensive documentation and a user-friendly interface for Gen AI developers.
    2. Implement real-time monitoring and auto-scaling of NIM services based on inference demand.
    3. Develop performance metrics to measure inference latency, throughput, and resource utilization across different model sizes and workloads.
  • Bonus Tasks:
    1. Demonstrate the use of your solution in a real-world Gen AI application, showing significant performance improvements with quantitative benchmarks.

Submission Format:

  • Workflow and technical documentation.
  • Source code on GitHub/GitLab.
  • Docker and Kubernetes deployment files.
  • Video demonstration of the Gen AI application's performance.

Resources:

  • NIM Documentation
  • Optimizing Transformer Model Inference
  • Scaling Generative AI in Production
Healthcare Innovation - Transform the Medical Landscape with Generative AI

Problem Statement Name: Advancing Healthcare with Multimodal Generative AI

Overview: Healthcare is poised for a revolution driven by Generative AI. Your challenge is to develop a multimodal AI-powered solution that significantly impacts the healthcare sector, from enhancing diagnostic processes to personalizing treatment plans and accelerating drug discovery. Leverage large language models, image generation, and other Gen AI technologies to create innovative solutions that improve patient outcomes, reduce costs, and advance the entire healthcare ecosystem.

Tech Stack:

  • Python, PyTorch, JAX
  • Hugging Face Transformers, Diffusers
  • Multimodal Generative AI Models (e.g., GPT-4 with vision, PaLM 2, MedPaLM)
  • Stable Diffusion or similar image generation models
  • Healthcare APIs and DICOM standard for medical imaging
  • Cloud Platforms (AWS HealthLake, Google Health Cloud, Azure Health Insights)

Restrictions:

  • Use at least one large language model and one image generation model tailored to healthcare applications.
  • Ensure adherence to healthcare industry standards and best practices

Exact Task:

  • Mandatory Tasks:
    1. Develop a multimodal AI model that aids in diagnosing medical conditions using both text and medical imaging data.
    2. Create a generative AI system that can produce synthetic medical data for research and training purposes.
    3. Implement a feature that uses Gen AI to personalize treatment plans based on patient data and medical literature.
  • Good-to-Have Tasks:
    1. Incorporate federated learning techniques to train your models while preserving patient data privacy.
    2. Develop a Gen AI-powered chatbot for patient engagement and preliminary symptom assessment.
    3. Implement metrics to evaluate the accuracy of AI-generated diagnoses, the quality of synthetic data, and the effectiveness of personalized treatment plans.
  • Bonus Tasks:
    1. Use Gen AI to accelerate drug discovery by generating and evaluating potential molecular structures, with quantitative metrics on the novelty and potential efficacy of generated molecules.

Submission Format:

  • Detailed workflow and technical documentation.
  • Source code hosted on GitHub/GitLab.
  • A video showcasing the multimodal AI solution in action.
  • PPT with a comprehensive overview, impact assessment, and future scope.

Resources:

  • Multimodal AI in Healthcare
  • Generative AI for Drug Discovery
  • Federated Learning in Medical AI

Prizes

Main Prizes
1st Prize
INR 5,00,000
2nd Prize
INR 4,00,000
3rd Prize
INR 3,00,000

Social Share

Help & Support

Please contact event admin
HackerEarth Support at support@hackerearth.com
Notifications
View All Notifications

?