I participated in my first Hackathon for Women in AI at Stanford University, organized by Twelve Labs, Zilliz, GenAI Collective, and Women Who Do Data (W2D2). One of the key insights I gained is that integrity is the most underrated value in today’s workplaces, often leading to conflicts and misunderstandings when overlooked.
In today’s fast-paced professional landscape, employees face numerous challenges, ranging from workplace stress to navigating career transitions. Despite the availability of HR teams, training programs, and mentorship opportunities, there remains a significant gap in providing real-time, personalized, and scalable guidance.
This introduces the Multi Agent Career and Workplace Assistant , an innovative application designed to address this gap. By leveraging AI-powered tools, it delivers personalized career coaching and workplace guidance, empowering employees to overcome challenges and achieve their professional goals.
The Problem
Employees often encounter challenges such as:
- Workplace Stress: Interpersonal conflicts, unclear communication, or work overload.
- Career Transition: Navigating skill gaps, identifying the right resources, and making informed career decisions.
Organizations face challenges too:
- Scalability: Providing mentorship and career coaching to all employees.
- Customization: Tailoring advice to individual needs.
- Engagement: Delivering relevant, on-demand learning resources.
The Solution: Multi-Agent Career and Workplace Assistant
This application uses a multi-agent framework to classify and address employee queries into two categories:
- Workplace Stress: Provides actionable advice and stress management resources.
- Career Transition: Generates a structured learning path for transitioning into new fields.
Key Features
- Intent Classification: Uses AI to determine whether the query is related to workplace stress or career transition
- Resource Retrieval: Embeds and retrieves relevant videos and PDFs using advanced embedding models such as Twelve labs’ Marengo Retriever 2.7 and Hugging Face Sentence transformer/all-MiniLM-L6-V2.
- Generative AI Integration: Delivers personalized advice using Gemini 2.0 Flash Experiment model.
- Scalability: Supports multiple users with minimal human intervention.
Technology Stack
- LangChain : For embedding and managing document embeddings.
- Twelve Labs : For video embeddings and processing.
- Zilliz/Milvus : To store and retrieve vector embeddings efficiently.
- Streamlit : For a user-friendly interface.
- Google Generative AI : To generate natural language responses using Gemini 2.0 Flash Experiment model.
- PyPDF2 : For PDF parsing and text extraction.
How does it Work?
Step 1: Intent Classification
The assistant classifies user queries into two categories:
- Career Transition: Queries about learning new skills or exploring new career paths.
- Workplace Stress: Queries about workplace conflicts, communication issues, or stress management.
Step 2: Resource Embedding and Retrieval
The application preprocesses and embeds PDFs and videos into Zilliz/Milvus, enabling fast and efficient similarity searches.
- Embedding PDFs:
- Embedding Videos:
Similarity Search:
Step 3: Creating System Prompt
Based on the classified intent, the assistant queries the embedding database for relevant resources and uses Generative AI to provide personalized recommendations.
Step 4: Streamlit UI
The application uses Streamlit for an intuitive UI where users input queries and receive personalized advice. The retrieved PDFs and videos are displayed with thumbnails and clickable links.
Conclusion
The Multi-Agent Career and Workplace Assistant bridges the gap between employees and scalable, personalized mentorship. By leveraging state-of-the-art AI tools, it provides timely and actionable guidance, ensuring both employees and organizations thrive in today’s dynamic professional environments.
Github: https://github.com/mrunmayee17/Career_and_Workplace_Assistant
Happy to connect on LinkedIn!
Top comments (0)