# GSoC 2025 Proposal - API Dash  
**Applicant**: Roshini D

---

## 📌 About

- **Full Name**: Roshini D  
- **Contact Info**: +91 9449761000  
- **Email**: drroshini16@gmail.com  
- **Discord**: rosh09068  
- **GitHub**: [roshcheeku](https://github.com/roshcheeku)  
- **LinkedIn**: [Roshini D](https://www.linkedin.com/in/roshini-d-94497525b?utm_source=share&utm_campaign=share_via&utm_content=profile&utm_medium=android_app)  
- **Time Zone**: GMT +5:30 (IST)  
- **Resume**: [View Resume](https://drive.google.com/file/d/19ZU_qnS_mRASstlQ0PDlReb52By-64Ls/view?usp=drive_link)

---

## 🎓 University Info

- **University**: BGS College of Engineering and Technology  
- **Program**: Bachelor of Engineering, Computer Science and Design  
- **Year**: Pre-final Year (3rd Year)  
- **Expected Graduation**: May 2026  

---

## 💡 Motivation & Past Experience

- **FOSS Contributions**:  
  Yes, contributed to GSSoC (Oct–Nov 2024):  
  - Tested API endpoints  
  - Implemented authentication systems  
  - Authored API documentation  

- **Proud Project**:  
  **Stock Sentiment Analysis and Prediction (95% accuracy)**  
  - Combined LSTM deep learning  
  - Integrated technical indicators and social media sentiment analysis  

- **Motivating Challenges**:  
  Real-world AI/ML problems with tangible impact, especially integrating multiple data sources and technologies  

- **Availability for GSoC**:  
  Will balance with Bachelor's program but can dedicate substantial time to GSoC  

- **Mentor Sync Preference**:  
  Yes, I welcome regular communication for project alignment  

- **Interest in API Dash**:  
  The mission of simplifying API evaluation resonates with my background in API testing and AI integration  

---

## 🔧 Ideas to Improve API Dash

- Automated batch testing  
- Version comparison functionality  
- Community benchmark repository  
- Framework adapters for easier integration  
- Include cost and latency metrics  
- A/B testing capabilities  

---

## 🧠 Project Proposal: AI API Evaluation Framework

A robust and extensible platform to evaluate and compare AI APIs from providers such as OpenAI, Google, Anthropic, and others.

---

### 📦 Core Components

- **Configuration Interface**  
  UI for input datasets, API credentials, request configs, and metrics  

- **Evaluation Engine**  
  Processes requests, collects responses, computes metrics  

- **Benchmark Repository**  
  Stores standard benchmarks for AI tasks, with support for custom benchmarks  

- **Visualization Dashboard**  
  Charts and tables for comparison results  

- **Batch Processing System**  
  Enables sequential/parallel evaluations  

---

### 🔧 Technical Implementation

- Adapters for major AI APIs  
- Flexible scoring system  
- Caching mechanisms  
- Exportable reports (PDF, CSV, JSON)  
- Extensible architecture  

---

### 👩‍💻 User Experience Focus

- Intuitive end-to-end workflow  
- Clear, comparative visualizations  
- Granular evaluation metric breakdowns  
- Easy sharing and exporting  

---

## 🗓️ Weekly Timeline

### Week 1–2: Research and Planning  
- Study existing frameworks and benchmarks  
- Define architecture and specs  
- Project setup and wireframes

### Week 3–4: Core Framework Development  
- Implement base evaluation engine  
- API adapter interfaces  
- Data models and testing infra

### Week 5–6: API Provider Integration  
- Adapters for OpenAI, Google, Anthropic  
- Auth handling, error/retry mechanisms  
- Test API connectivity and parsing

### Week 7–8: Benchmark Implementation  
- Text and sentiment benchmarks  
- Image-based benchmarks  
- Interface for custom benchmarks

### Week 9–10: UI Development  
- Config interface  
- Dashboard visualizations  
- Batch UI and settings

### Week 11–12: Visualization and Reporting  
- Comparison graphs  
- Drill-down result views  
- Export and share features

### Week 13–14: Testing and Optimization  
- Test across APIs and benchmarks  
- Performance tuning and caching  
- Bug fixing

### Week 15: Final Polish  
- Write user & developer docs  
- UI polish and final optimizations  
- Final presentation and submission

---