IC-Easy

Inspiration

As students and researchers ourselves, we experienced firsthand the frustration of juggling multiple tools and websites throughout the academic writing process. Writing a research paper meant switching between literature search engines to find relevant papers, grammar checkers to correct our writing, citation generators to format references, plagiarism detectors to ensure originality, note-taking apps to organize our findings, and separate tools to visualize research connections.

This fragmented workflow was not only time-consuming but also broke our concentration and creative flow. We envisioned a unified platform where a student could go from a research keyword to a publication-ready paper with proper citations and originality verification, all in one place, powered by Chrome's on-device AI for privacy and offline access.

What it does

IC-Easy is an all-in-one academic research and writing platform that transforms the entire research paper workflow into a seamless experience. The platform combines literature discovery, AI-powered analysis, intelligent writing assistance, and quality assurance into a single integrated environment.

Literature Discovery & Management Users can search across multiple academic databases (CrossRef, arXiv, OpenAlex) simultaneously from a single search bar. The system automatically deduplicates results, enriches metadata, and presents papers with rich preview cards including citations, publication year, and abstracts. Users can save papers to personalized collections and track their research history.

AI-Powered Summarization Using Chrome's Summarizer API, IC-Easy generates concise summaries of complex research papers entirely on-device. Users can quickly extract key findings from multiple papers without reading each in full. The hierarchical summarization approach handles papers of any length by processing sections independently and synthesizing an overall summary.

Intelligent Argument Extraction The Prompt API analyzes summarized papers to identify and extract key arguments, methodologies, and findings. This structured extraction helps users understand the core contributions of each paper and how they relate to their research question.

Academic Writing Assistant IC-Easy's writing engine combines multiple Chrome AI APIs to generate publication-quality academic content. The Writer API creates original drafts incorporating user-selected research points, the Rewriter API refines content for better flow and clarity, and the Proofreader API ensures grammatical accuracy. All content generation happens on-device, protecting user privacy and enabling offline work.

Citation Management The platform automatically formats references in APA, IEEE, and MLA styles. Users can export citations in BibTeX and RIS formats for use with other reference managers. The citation formatter parses paper metadata and applies style-specific rules for author ordering, title capitalization, and punctuation.

Plagiarism Detection IC-Easy includes an originality checker that compares user-written content against source papers using embedding-based similarity analysis. The system also performs statistical analysis to detect potential AI-generated content patterns, providing students with feedback about their work's originality.

Knowledge Graph Visualization The platform generates interactive citation networks and co-author relationship graphs using D3.js. Users can explore how papers connect through citations, discover influential works in their field, and identify key researchers. The visualization includes depth-limited traversal to maintain performance even with highly-cited papers.

How we built it

Architecture Overview

Our application follows a hybrid client-server architecture that maximizes Chrome AI capabilities while leveraging backend services where necessary.

Frontend Layer

Built with Next.js and React, providing a responsive interface with real-time feedback. TailwindCSS handles styling while D3.js powers interactive knowledge graph visualizations. The frontend directly interfaces with Chrome Built-in AI APIs through browser-native calls, ensuring all content processing happens on-device.

Chrome AI Integration

We implemented multiple AI workflows that run entirely client-side:

Literature summarization using the Summarizer API to distill abstracts and full papers into concise insights
Argument extraction through carefully crafted prompts using the Prompt API to identify key research contributions
Draft generation with the Writer API, incorporating user-selected research points and maintaining academic tone
Content refinement with the Rewriter API for alternative phrasings and improved flow
Grammar correction via the Proofreader API as a final polish before export

Backend Services

Python FastAPI server handles computationally intensive tasks that benefit from centralized processing:

Multi-source literature search across CrossRef, arXiv, and OpenAlex with parallel query execution
Citation graph construction by traversing reference networks and computing relevance scores
Plagiarism detection using text embedding similarity comparisons and statistical pattern analysis
User authentication and search history persistence with JWT-based security
Paper metadata enrichment and normalization across heterogeneous data sources

Data Flow Architecture

When a user searches for literature, the backend aggregates results from multiple academic databases, deduplicates entries by DOI and normalized titles, then returns structured data to the frontend. The frontend displays papers with rich metadata, and users can select papers to summarize using Chrome AI. Selected papers and their summaries feed into the Writer API to generate coherent drafts incorporating the research. The generated text goes through Rewriter and Proofreader APIs before presenting the final output to users.

Key Technical Decisions

Client-side Processing Prioritization

We route all content generation, summarization, and text refinement through Chrome AI rather than cloud APIs. This architectural decision ensures user data never leaves their device, provides privacy guarantees, and enables offline writing capability after initial paper collection. Students can work on their papers anywhere without internet connectivity concerns.

Modular API Composition

Rather than treating Chrome AI as a monolithic service, we composed a pipeline where each API handles its specialized task. Generating an academic paper involves a coordinated sequence: Summarizer condenses source papers, Prompt API extracts key thesis statements and arguments, Writer API drafts sections incorporating these insights, Rewriter API improves content flow and coherence, and Proofreader API polishes the final text. Each stage validates its input and output for seamless transitions.

Progressive Enhancement Strategy

The application works at multiple capability levels. Basic search and collection functionality works without Chrome AI, providing value even in unsupported browsers. Summarization adds significant value when Chrome AI is available, helping users quickly process multiple papers. Full writing assistance requires the complete API suite but delivers the maximum benefit. This approach ensures broad accessibility while encouraging users to upgrade to supported environments.

Reference Formatting Engine

We implemented a template-based citation formatting system supporting APA, IEEE, and MLA styles. The system parses paper metadata including authors, titles, journals, publication years, and DOIs, then applies style-specific rules for author ordering, title capitalization, punctuation, and formatting. The engine handles edge cases like missing data, multiple authors, and corporate authors correctly.

Intelligent Chunking for Long Papers

To handle research papers exceeding token limits, we implemented hierarchical processing that splits papers at natural boundaries like section headers and paragraph breaks rather than arbitrary token counts. Each section is processed independently, then results are synthesized into coherent overall summaries and analyses. This approach preserves paper structure and context while respecting API constraints.

Challenges we ran into

Orchestrating Multiple AI APIs

Each Chrome AI API has different capabilities, token limits, and processing times. Designing a coherent workflow that chains these APIs together without creating a disjointed user experience proved complex. We needed to ensure outputs from one API were compatible with inputs to the next while providing clear feedback to users during multi-step processing.

We implemented a state machine approach where each stage validates its input and output, ensuring compatibility with the next stage. For example, the Summarizer output is validated for length and structure before feeding into the Prompt API for argument extraction. We added loading states and progress indicators so users understand the multi-step processing happening on their device.

Context Window Management

Research papers can be lengthy, often exceeding Chrome AI's token limits by several multiples. We needed strategies to handle papers that are 10,000 words or more while maintaining coherent summaries and analysis that capture the full paper's contributions.

We implemented intelligent chunking that splits papers at natural boundaries such as section headers and paragraph breaks rather than arbitrary token counts. For summarization, we process each section independently, then use the Prompt API to synthesize an overall summary that captures the paper's main contributions. This hierarchical approach preserves paper structure and maintains context while respecting token limits.

Integrating Heterogeneous Data Sources

CrossRef, arXiv, and OpenAlex return data in wildly different formats. CrossRef uses DOIs and structured metadata with specific field names, arXiv uses its own ID system with different metadata schemas, and OpenAlex has yet another structure with unique identifiers. Normalizing these into a consistent format while preserving source-specific features and maintaining data quality was challenging.

We created transformation layers that map each source to a unified internal schema with optional fields for source-specific data. Deduplication logic checks both DOI and normalized titles to merge duplicates across sources, handling cases where the same paper appears in multiple databases with slight variations. The frontend receives consistent data regardless of origin, simplifying the UI layer.

Balancing On-device and Server-side Processing

Determining which features should run through Chrome AI versus backend services required careful consideration of multiple factors including privacy, performance, offline capability, and resource requirements. We wanted to maximize privacy and offline capability while acknowledging that some tasks genuinely benefit from server-side processing.

We established clear criteria based on data requirements and privacy implications. Content generation, summarization, and refinement stay on-device for privacy protection and offline access. Literature search, citation graphs, and plagiarism detection use backend services because they require access to large external datasets and benefit from centralized caching. User preferences and history sync through the backend for cross-device access while keeping research content strictly local.

Knowledge Graph Performance

Citation networks can become massive very quickly. A single popular paper might cite hundreds of works and be cited by thousands more, creating networks with tens of thousands of nodes. Rendering these graphs in real-time while maintaining interactivity and preventing UI freezing required careful optimization at both the data and visualization layers.

We implemented depth-limited traversal with configurable maximum node counts to prevent exponential explosion. The backend pre-processes graph structure and returns only the most relevant connections based on citation counts and recency. The frontend uses force-directed layout algorithms with collision detection to prevent node overlap and enable smooth interaction. We also implemented lazy loading so only visible portions of large graphs are fully rendered.

Plagiarism Detection Accuracy

Traditional plagiarism detection compares against massive known databases and uses sophisticated linguistic analysis. We needed to detect potential plagiarism and AI-generated content while measuring originality against the user's source papers, all without access to commercial detection services or large comparison databases.

We implemented a two-pronged approach combining multiple detection strategies. First, embedding-based similarity comparison against source papers detects excessive copying or paraphrasing. Second, statistical analysis of writing patterns including vocabulary diversity, sentence structure variation, and transition usage flags potential AI generation. While not as comprehensive as commercial tools, this approach provides meaningful feedback to students about their work's originality and helps them improve their writing.

Accomplishments that we're proud of

Seamless Chrome AI Integration

We successfully orchestrated all five Chrome Built-in AI APIs (Summarizer, Prompt, Writer, Rewriter, Proofreader) into a coherent pipeline that feels natural to users. The multi-API workflow processes academic content through several stages of refinement while providing clear feedback at each step. This represents one of the most comprehensive uses of Chrome's AI capabilities in a single application.

Privacy-First Architecture

By routing all content generation and analysis through on-device Chrome AI, we created a platform where student research and writing never leaves their device. This privacy guarantee is particularly important for academic work, where students may be working with sensitive research topics or unpublished ideas. The offline capability also makes the tool accessible to students with unreliable internet access.

Unified Research Workflow

We consolidated what typically requires five to seven separate tools into a single integrated platform. Students can now go from initial literature search to publication-ready paper with formatted citations and originality checking without switching contexts. This reduces cognitive load and allows students to maintain focus on their research and writing.

Real-time Knowledge Graphs

The interactive citation network visualization provides immediate insight into research landscapes that would typically require hours of manual exploration. Students can visually understand how papers relate, identify seminal works, and discover new research directions. The performance optimizations allow smooth interaction even with complex networks.

Intelligent Context Management

Our hierarchical processing approach successfully handles papers of any length while respecting Chrome AI's token limits. The system intelligently splits content, processes sections independently, and synthesizes coherent results that capture the full document's meaning. This makes the tool practical for real-world academic papers.

Cross-Database Search

Aggregating results from CrossRef, arXiv, and OpenAlex provides comprehensive coverage of academic literature. The deduplication and normalization logic ensures students see relevant papers once with consistent metadata, regardless of which databases contain them. This saves time and reduces confusion from duplicate results.

What we learned

Chrome Built-in AI Capabilities and Constraints

We explored the full suite of Chrome AI capabilities, learning how to orchestrate the Summarizer API for distilling complex papers, the Prompt API for extracting key arguments, the Writer API for generating coherent content, the Rewriter API for refining drafts, and the Proofreader API for ensuring grammatical accuracy. Understanding when to use each API and how to chain them together was crucial for creating natural workflows.

We also learned the importance of designing for client-side processing constraints including token limits, context windows, and processing speed. Unlike cloud-based solutions where we could throw unlimited compute at problems, Chrome AI required us to be strategic about what we process locally and how we structure that processing.

On-device AI Paradigm Shift

Working with Chrome's on-device AI taught us to think differently about application architecture. Traditional web apps assume unlimited backend resources and constant connectivity. Chrome AI applications must be designed for resource-constrained environments with intermittent connectivity while still delivering powerful features. This paradigm shift influenced every architectural decision we made.

Academic Data Integration Challenges

Working with CrossRef, arXiv, and OpenAlex APIs taught us how heterogeneous academic databases are. Each has different data structures, citation formats, and query patterns. We learned to normalize and deduplicate results across sources while preserving metadata integrity. Understanding the quirks of each API and handling edge cases like missing DOIs, inconsistent author names, and varied date formats was essential.

Knowledge Graph Construction

Implementing citation networks and co-author relationships required understanding graph algorithms, data structures, and visualization techniques. We learned to balance between comprehensive data collection and performance, as citation networks can explode exponentially if not properly bounded. Choosing which relationships to include and how to weight connections significantly impacts both performance and usefulness.

Hybrid Architecture Benefits

We discovered that the most powerful solution combines Chrome's on-device AI with strategic backend services. Certain tasks like literature search and plagiarism detection benefit from centralized processing, while summarization and writing work beautifully on-device. Learning to identify which features belong where and designing clean interfaces between layers was a key insight.

User Experience for Multi-step AI Workflows

Chaining multiple AI APIs creates complex workflows that can confuse users if not carefully designed. We learned the importance of progress indicators, intermediate feedback, and clear state transitions. Users need to understand what's happening during processing, how long it might take, and what the results mean. Good UX for AI features requires transparent communication about what the AI is doing.

What's next for IC-Easy

Enhanced Collaboration Features

We plan to add real-time collaboration capabilities allowing research teams to work together on papers. Multiple users could contribute to shared literature collections, annotate papers collaboratively, and co-write sections with AI assistance. Version control would track contributions and changes, while Chrome AI ensures each user's device handles their content privately.

Semantic Similarity Clustering

Beyond citation-based relationships, we want to implement semantic similarity analysis that clusters papers by conceptual themes. This would help researchers discover relevant papers that may not cite each other but address similar questions or use related methodologies. The clustering visualization would reveal conceptual landscapes of research fields.

Multilingual Support with Translator API

The Chrome Translator API could enable cross-language research capabilities. Students could search literature in any language and have abstracts automatically translated, making global research more accessible. The Writer API could also generate papers in multiple languages, helping students whose native language differs from their publication target.

Advanced Writing Assistance

We envision more sophisticated writing features including tone adjustment for different academic styles, automatic section structuring based on paper type (research article, review, thesis chapter), and context-aware suggestions that reference specific papers in the user's collection. The AI could suggest where to cite particular papers based on content relevance.

Mobile Application

Extending IC-Easy to mobile devices would allow students to conduct literature reviews and work on papers anywhere. A React Native mobile app leveraging Chrome AI on Android (via WebView with AI support) would provide the full feature set on smartphones and tablets. Offline-first design would make the mobile experience reliable even with spotty connectivity.

Integration with Reference Managers

Direct integration with popular reference managers like Zotero, Mendeley, and EndNote would allow users to import existing libraries and export results. This would make IC-Easy fit naturally into established academic workflows rather than requiring users to abandon their existing tools.

Citation Recommendation Engine

An intelligent citation recommendation system could analyze a user's draft and suggest relevant papers they should cite based on their arguments and claims. This would help ensure comprehensive literature coverage and reduce the risk of missing important related work.

Institutional Repository Integration

For universities using IC-Easy, we could integrate with institutional repositories to enable one-click submission of finished papers. The platform could also highlight research from the user's institution, fostering awareness of local expertise and potential collaboration opportunities.

Advanced Plagiarism Detection

Expanding plagiarism detection capabilities with more sophisticated linguistic analysis, integration with academic integrity databases, and more nuanced reporting would make IC-Easy a comprehensive originality checking solution. We could also add features to help students understand how to properly paraphrase and cite sources.

Research Trend Analysis

Using topic evolution tracking across years, we could show students how research areas are developing, which topics are gaining attention, and where gaps might exist. This meta-analysis would help students identify promising research directions and understand the broader context of their work.