Mark4K

This project is built for digital ocen 4k to help students make better carrier path in school

Comment

Inspiration

University life is overwhelming. Between crushing assignment deadlines, the anxiety of choosing the right career path, and the pressure to maintain professional communication, many students feel lost. We realized that students don't just need a search engine; they need a mentor.

The inspiration for Mark4K came from the desire to make the path from "Student" to "Professional" easier. We wanted to build a tool that doesn't just answer questions, but actively helps students execute tasks—whether that’s planning a PhD trajectory, debugging code for a final year project, or drafting a polite email to a strict professor. We built Mark4K to be the 24/7 academic partner we wish we had.

What it does

Mark4K is an all-in-one AI Academic Assistant powered by DigitalOcean Gradient AI. It operates in five distinct modes:

  1. Chat Advisor: A general-purpose academic mentor for brainstorming and advice.
  2. Education Path: A career counselor that maps out the journey from BSc to PhD to Professor, including scholarship and university recommendations.
  3. PDF Analyzer: Students can upload textbooks, research papers, or assignment sheets. Mark4K reads them, summarizes key points, and extracts requirements.
  4. Code Generator: Generates full-stack code (Python, HTML, JS) for CS projects and packages them into a downloadable ZIP file.
  5. Email Drafter: Drafts professional emails to lecturers and sends them directly via Gmail SMTP integration.

How we built it

  • Backend: We used Python and Flask to handle the server-side logic and session management.
  • AI Engine: The core intelligence is powered by DigitalOcean Gradient AI (using the Qwen3 32B model). We engineered specific system prompts to make the agent switch personalities based on the selected "Mode."
  • Frontend: A single-page application (SPA) built with vanilla HTML/CSS/JS, designed with a "Dark Academic" aesthetic (deep navy and gold) to reduce eye strain during late-night study sessions.
  • Libraries:
    • pypdf for extracting text from uploaded assignment documents.
    • fpdf2 for generating downloadable project plans and summaries.
    • smtplib for the integrated email sender.

Challenges we ran into

The biggest technical hurdle was PDF generation and analysis.

  1. The "PDF Failed" Error: We initially struggled with fpdf2 compatibility. Generating a PDF on the fly from AI-generated Markdown text—while handling encoding issues and binary output streams—caused the app to crash frequently. Debugging the binary flow between the Flask backend and the browser download trigger was a significant challenge.
  2. Token Limits: Uploading large textbooks often exceeded the context window of the AI model. We had to implement logic to truncate and sanitize the extracted PDF text before sending it to the Agent.
  3. Prompt Engineering: Getting the AI to switch from being a "Python Coder" to a "Career Counselor" required careful tuning of the conversation history and system prompts.

Accomplishments that we're proud of

  • Seamless PDF Integration: We successfully fixed the PDF engine! Users can now not only upload documents for analysis but also download the AI's advice as a branded, professional PDF file.
  • Instant Code Deployments: The "Download ZIP" feature is a game-changer. The AI doesn't just write code snippets; it packages a full project structure that a student can unzip and run immediately.
  • The UI/UX: We are proud of the "Dark Academic" design. It feels premium and focused, unlike the cluttered interfaces of standard chatbots.

What we learned

  • Library Dependencies matter: We learned the hard way that ensuring the correct versions of libraries like fpdf2 and pypdf is crucial for binary file handling.
  • The power of Context: We learned that maintaining a "conversation history" (memory) significantly improves the quality of the AI's advice, but we had to balance that against browser cookie limits.
  • Students need practical tools: It wasn't enough to just "chat." Adding features like Email Sending and ZIP downloading turned this from a toy into a utility.

What's next for Mark4K

  • Scaling with DigitalOcean: We aim to move from a local Flask app to a scalable deployment using DigitalOcean App Platform to handle thousands of concurrent student users.
  • University Collaboration: We want to partner with software organizations and educational institutions to integrate university-specific syllabi directly into the AI's knowledge base.
  • Voice Interface: Adding speech-to-text so students can rehearse their thesis defenses with Mark4K.
  • Real-time Collaboration: Allowing groups of students to work in a shared Mark4K session for group projects.

Built With

  • digitalocean-gradient
Share this project:

Updates