AI Powered Ed-TECH Platform
Aayush Pagare
Software Engineer
Web Developer
Technical Writer
LangChain
Next.js
PostgreSQL
Aspirants AI: Empowering UPSC Aspirants with AI-driven tools.
AspirantsAI is an all-in-one platform that leverages artificial intelligence to enhance the preparation process for UPSC Mains aspirants. It features AffairsQuest, an intelligent tool that connects current affairs to relevant PYQs, and SmartCheck, which provides detailed feedback on handwritten answers. Designed to streamline studies, AspirantsAI empowers aspirants with efficient search capabilities, insightful answer reviews, and tools for comprehensive preparation. Whether you're aiming to master current affairs, improve answer-writing skills, or access tailored resources, AspirantsAI is your companion for UPSC success.
Smartcheck
SmartCheck is an AI-driven tool designed to enhance the UPSC Mains preparation process by providing personalized evaluations for handwritten answers. Aspirants can upload their PDFs, and SmartCheck's advanced AI analyzes key aspects such as content, accuracy, structure, clarity, and depth. The tool offers valuable, actionable feedback to help users improve their answer-writing skills and boost their exam performance. SmartCheck is an indispensable companion for UPSC candidates, delivering expert-level evaluations that allow for smarter, more efficient preparation.
SmartCheck operates through a sophisticated, multi-step process to evaluate UPSC Mains answer scripts with remarkable accuracy:
OCR Text Extraction with AWS Textract:
The process begins with AWS Textract, a powerful Optical Character Recognition (OCR) tool, which scans the handwritten PDF of the answer script.
Textract excels in converting even complex, less legible handwriting into accurate digital text. Its advanced capabilities ensure that no important content is lost, even when dealing with intricate handwriting styles or unclear text.
Analysis with Google GeminiAI:
After the text is ready, it is passed to Google GeminiAI, a state-of-the-art generative AI system designed to perform sophisticated natural language understanding and analysis.
GeminiAI uses its deep learning models to understand the context, structure, and overall quality of the answer. It evaluates how well the answer aligns with the question, its logical flow, and the depth of insight provided.
Scoring with a Tailored Rubric:
GeminiAI then applies a structured scoring rubric that has been specifically developed for UPSC answer evaluation. This rubric assesses multiple key parameters such as:
Content Relevance: Whether the answer addresses all aspects of the question and is factually accurate.
Structural Coherence: How logically the answer is organized, including the flow of ideas and clarity of the introduction, body, and conclusion.
Argument Clarity: The clarity and persuasiveness of the arguments, including how well-supported and detailed the reasoning is.
and many more features
Evaluation Results:
Once the evaluation is complete, the system produces a detailed report, offering a score for each parameter and an overall evaluation of the answer. This score reflects how well the answer matches UPSC’s demanding standards.
The results are then presented to the user in a clear, actionable format, allowing them to understand their strengths and areas for improvement.
This seamless integration of OCR technology, generative AI analysis, and a custom scoring rubric empowers SmartCheck to provide UPSC aspirants with valuable, AI-driven feedback on their handwritten answers.
Please check out detailed blog here: Medium-Blog
Affairs Quest
AffairsQuest follows a multi-step process to link current affairs articles to relevant UPSC Mains previous year questions (PYQs):
Storing PYQs in PostgreSQL:
All previous year questions (PYQs) are stored in a PostgreSQL database. The questions are organized in a structured manner, likely with columns for the question, exam year, and topic/category.
This allows efficient storage and querying of the PYQs based on different criteria (such as year, topic, etc.).
Generating Vector Embeddings:
Google Generative AI embeddings are used to create vector representations for both the PYQs and articles. The embeddings are stored in a Neon PostgreSQL database.
Embeddings allow for semantic similarity matching. Each article and PYQ is transformed into a high-dimensional vector representation that captures its semantic meaning.
Parsing and Cleaning the Article:
Articles from The Hindu are parsed. This likely involves web scraping, using a tool like
Cherrio to extract the article content, cleaning it by removing HTML tags, special characters, or irrelevant text, and ensuring that only the main body of the article remains.The parsed text is then preprocessed (tokenization, lowercasing, removing stop words, etc.) to ensure it is in an optimal format for generating embeddings.
Generating Article Embeddings:
After cleaning, the article text is fed into Google Generative AI’s embedding model. This model produces a vector that represents the content of the article in a way that can be compared to the PYQs’ embeddings.
Similarity Search:
A similarity search is performed by comparing the article’s embedding with the embeddings of stored PYQs in the database.
The system retrieves the most similar PYQs based on the cosine similarity between their embeddings. This step allows users to see which past questions are most relevant to the current article.
Serving Results:
The retrieved PYQs are returned to the user, often along with a relevance score based on the similarity ranking.
The system may display the results directly in the interface or use them to generate insights about which areas of the article are most aligned with the exam’s focus.
This architecture leverages a combination of web scraping, text preprocessing, vector embeddings, and similarity search to connect current affairs with the UPSC Mains exam syllabus, providing a useful tool for aspirants.
In conclusion, AspirantsAI continues to evolve, and the next phase involves exploring new tools to further enhance the experience for UPSC aspirants. One such tool is NoteTalk, which enables users to interact with their handwritten notes through a conversational interface, allowing for dynamic and insightful engagement with their study materials. Additionally, the integration of hybrid search combined with semantic search will ensure more efficient and relevant information retrieval, blending traditional keyword-based searches with advanced AI-driven understanding.
To contribute: Github
