[DMP 2024]:

[piyushmittaltech]

Ticket Contents

Protean Overview

Protean eGov Technologies Limited provides IT services. The Company offers citizen services, e-governance solutions, system integration, business process re-engineering, data center colocation, and IT consulting services for citizens, corporations and the Government. We have been modernizing the direct tax infrastructure, providing a tax identity to citizens and corporates (issuance of PAN card), strengthening the old-age social security infrastructure (National Pension System NPS & Atal Pension Yojna - APY), promoting financial inclusion by contributing to the India Stack by enrolling citizens for National Identity and enabling the BFSI sector by providing Aadhaar-based identity authentication and e-Sign services

Benefits of this Project

• Improved search functionality with AI/ML-based algorithms, leading to better user experience and engagement.
• Enhanced inventory management through accurate forecasting, resulting in optimized inventory levels and reduced costs.
• Streamlined development process with well-documented APIs, facilitating easier integration and maintenance.

Description

Develop a set of APIs containing AI/ML-based search algorithm integration and an Inventory Forecasting System to provide efficient search capabilities and accurate inventory predictions. The APIs should be well-documented using Swagger Documentation and thoroughly tested to ensure reliability and performance.

AI/ML-based Search Algorithm API Integration: As a developer, I want to integrate an AI/ML-based search algorithm API into our system. Tasks:

• [ ] Research and select suitable AI/ML search algorithm API.
• [ ] Implement API integration into our system.
• [ ] Test API functionality for accuracy and performance.

Inventory Forecasting API Development: As a developer, I want to create an API for inventory forecasting. Tasks:

• [ ] Design API endpoints for inventory forecasting.
• [ ] Develop AI/ML models for forecasting.
• [ ] Integrate forecasting models into API endpoints.
• [ ] Test API functionality with sample data.

Goals & Mid-Point Milestone

Mid-Point Goals:

AI/ML-based Search Algorithm API Integration:

• [ ] Completion of AI/ML search algorithm API integration with basic search functionality.
• [ ] Testing of API endpoints for initial accuracy and performance.

Inventory Forecasting API Development:

• [ ] Successful development and integration of API endpoints for inventory forecasting.
• [ ] Testing of forecasting API with sample data to ensure accuracy.

Final Goals:

AI/ML-based Search Algorithm API Integration:

• [ ] Full integration of AI/ML-based search algorithm API with advanced features such as natural language processing.
• [ ] Testing of API endpoints with diverse datasets for accuracy and performance optimization.
• [ ] Documentation of API endpoints using Swagger.

Inventory Forecasting API Development:

• [ ] Implementation of advanced features in the inventory forecasting API, such as real-time updates and adaptive algorithms.
• [ ] Comprehensive testing of forecasting API with data to validate accuracy and reliability.
• [ ] Documentation of API endpoints using Swagger.

Setup/Installation

No response

Expected Outcome

APIs with Advanced Search Functionality:

• APIs integrated with AI/ML-based search algorithm offering accurate and efficient product search capabilities.
• Swagger documentation providing clear specifications and usage instructions for API endpoints.

Inventory Forecasting APIs for Efficient Inventory Management:

• APIs for inventory forecasting integrated into the system, providing accurate predictions for optimal inventory management.
• Swagger documentation offering comprehensive documentation of API endpoints, including request and response structures.

Acceptance Criteria

AI/ML-based Search Algorithm API Integration:

• [ ] The API should return relevant results with high accuracy.
• [ ] API endpoints should be responsive and efficient, providing results in real-time.
• [ ] Swagger documentation should be comprehensive, detailing each endpoint's functionality and usage.

Inventory Forecasting API Development:

• [ ] The forecasting API should provide accurate predictions for inventory levels over a specified period.
• [ ] Forecasting API endpoints should be well-documented in Swagger, including request parameters and response structures.
• [ ] The system should be able to adapt to changes in demand and provide timely updates for inventory management decisions.

Implementation Details

Go / Node.JS / Python (any one technology to be used)

Mockups/Wireframes

Not Applicable

Product Name

Seller App

Organisation Name

Protean

Domain

eCommerce

Tech Skills Needed

Database, Machine Learning, Python, Swagger

Mentor(s)

@piyushmittaltech @RaviV-Protean @amitknamdeo @sunilnsdlekyc

Category

API, Machine Learning

[Sushila-beniwal]

Hi @piyushmittaltech,@RaviV-Protean, @amitknamdeo ,@sunilnsdlekyc I'm very excited to apply my skills(machine learning , data scrapping ,data analysis, deep learning) to solve real life problem. And i think it would be a great opportunity for me to work with your team it would help me a lot in my learning phase. I would like to connect with you. Awaiting for your reply sir..... Thank you.

[Janhavibabber]

Hi @piyushmittaltech, @RaviV-Protean, @amitknamdeo, @Sushila-beniwal

I hope this message finds you well. I wanted to express my keen interest in contributing to the development project for Protean. With my background in [AI/ML-based Algorithm, building APIs, python]. I am confident in my ability to make meaningful contributions to this initiative.

I am excited about the prospect of leveraging my skills to help Protean improve search functionality with AI/ML-based algorithms, leading to better user experience and engagement and enhanced inventory management through accurate forecasting, resulting in optimized inventory levels and reduced costs.

I am eager to collaborate with the team and contribute to the success of this project. Also interested to know about the tools being used by Protean for this particular development.

Looking forward to the opportunity to work together and make a positive impact.

[furiousluck]

Hi @piyushmittaltech , @RaviV-Protean, @amitknamdeo, @sunilnsdlekyc I have previously worked on a similar project that involved developing an AI/ML model to predict outcomes for specific conditions and integrating it with an API built using Node.js. I believe this experience positions me well to contribute effectively to this project. I am enthusiastic about the opportunity to collaborate with you on it. Thanks

[Sufia-ahmad]

sir, I want to do this project from my previous experiences.

[ashuashutosh2211]

Hi @piyushmittaltech, @RaviV-Protean, @amitknamdeo, @sunilnsdlekyc,

I'm thrilled to apply my skills in machine learning, data scraping, data analysis, deep learning, and problem-solving to address real-life challenges. I have done some proejcts related to machine learning and deep learning also such as Stock Price Prediction, Voice Controlled Music Recommendation System (in this project I have applied recommendation algorithms using NLP which are similar to Search algorithm) . Joining your team would be a valuable learning experience for me. I'm eager to connect with you. Waiting for your reply. Thank you!

[AbhimanyuSamagra]

Do not ask process related questions about how to apply and who to contact in the above ticket. The only questions allowed are about technical aspects of the project itself. If you want help with the process, you can refer instructions listed on Unstop and any further queries can be taken up on our Discord channel titled DMP queries. Here's a Video Tutorial on how to submit a proposal for a project.

[Jashvinu]

Hello @piyushmittaltech @RaviV-Protean @amitknamdeo @sunilnsdlekyc

How does the current search system determine search result ranking? Could the AI/ML model provide signals to improve relevance scoring and are there ways to potentially personalise the ranking for users?

Also given the potential seasonality of demand, would you lean towards time-series models ( ARIMA, SARIMA) or are there advantages to tree-based ensembles (XGBoost) or something that I have personally been working on is Lag LLama Zero-Shot LLM I think that would be worth exploring?

[HYDRAEZIO]

Hello @piyushmittaltech, @RaviV-Protean, @amitknamdeo, and @sunilnsdlekyc,

I hope this message finds you well. I'm excited about the opportunity to work on the "Seller App" project with Protean and contribute my skills and experience to its success.

Regarding the project details you've provided, I find it quite intriguing and aligning with my expertise. Here's why I believe I would be a valuable addition to your team:

I have participated in the Smart India Hackathon, where I worked on a Digilocker function known as ABC ID. Our team successfully developed and integrated this functionality, contributing to digital identity management solutions. This experience has honed my skills in understanding government requirements, security standards, and the importance of reliable and scalable solutions in the e-governance domain.

I am eager to contribute to the development of advanced search functionality using AI/ML algorithms and the inventory forecasting system for the Seller App. The goals and milestones outlined in the proposal are well-defined, and I am confident in my ability to meet and exceed these expectations.

Looking forward to the opportunity to collaborate with Protean and deliver a robust and innovative solution.

[TanmayAT]

Hi , @piyushmittaltech

Certainly! Here's a tailored markdown message specifically for the Protean eGov Technologies Limited project:

---

Approach to Developing AI/ML-based Search Algorithm Integration and Inventory Forecasting APIs for Protean eGov Technologies Limited

1. Understanding the Requirements:

We'll start by thoroughly understanding the project requirements provided by Protean eGov Technologies Limited. This involves identifying the specific needs for enhancing search functionality and improving inventory forecasting within the e-governance and citizen services domain.

2. Research and Selection of AI/ML Technologies:

We'll conduct in-depth research to identify AI/ML technologies suitable for enhancing search capabilities. This includes considering factors such as natural language processing (NLP) for advanced search queries and machine learning algorithms for accurate inventory forecasting.

3. Designing API Endpoints:

Based on the requirements gathered, we'll design clear and intuitive API endpoints for both the AI/ML-based search algorithm integration and the inventory forecasting system. These endpoints will outline the functionalities and data interactions required by the APIs.

4. Development and Integration:

Our development team will implement the selected AI/ML technologies to integrate advanced search algorithms into the existing system. Simultaneously, we'll develop machine learning models for inventory forecasting and seamlessly integrate them into the API endpoints.

5. Testing and Validation:

Thorough testing will be conducted to validate the functionality, accuracy, and performance of both APIs. This includes rigorous testing with sample data to ensure that search results are relevant and inventory forecasts are reliable.

6. Documentation:

We'll document the APIs meticulously using Swagger Documentation. This documentation will provide comprehensive details about each API endpoint, including parameters, expected responses, and usage examples. Clear documentation is crucial for easy integration and maintenance by Protean eGov Technologies Limited.

7. Iterative Optimization:

We'll continuously iterate on the development and optimization of both APIs based on feedback and performance metrics. This may involve fine-tuning the AI/ML models, enhancing search relevance, improving forecasting accuracy, and adding new features as needed.

8. Deployment and Monitoring:

Upon completion, we'll deploy the APIs into the production environment of Protean eGov Technologies Limited. We'll implement monitoring mechanisms to track performance metrics and gather user feedback. Regular reviews and updates will ensure that the APIs remain effective and aligned with the organization's goals.

By following this tailored approach, we'll develop and deploy AI/ML-based search algorithm integration and inventory forecasting APIs that specifically cater to the needs of Protean eGov Technologies Limited, contributing to the modernization of e-governance and citizen services.

---

[Sumitkevlani]

Hi @piyushmittaltech, here is my approach of working on the project

Approach for AI/ML-based Search Algorithm API Integration:

Gather Requirements:

We will ensure responsiveness and efficiency by understanding the requirements for the search functionality and determining the types of data to be searched and user preferences.

Select NLP-Based Model:

We will leverage Natural Language Processing (NLP) algorithms for search functionality, preprocessing data to generate searchable tags, ensuring accurate search results.

Model Fine-Tuning:

To optimize performance, we will conduct hyperparameter tuning, experimenting with different Stemming approaches, Scaling techniques, and Vectorization methods, evaluating each model's effectiveness.

Develop API:

Utilizing the selected NLP model, we will craft the search API, aiming for seamless integration and intuitive user experience.

Testing:

To validate accuracy and performance, we will thoroughly test API functionality with sample data, ensuring it meets or exceeds expected outcomes.

Alternative Approach:

If necessary, we will explore generative AI and foundational models, using LangChain, to enhance search capabilities, adapting our approach as needed to deliver optimal results.

Additional Feature:

Intrigued by the potential for user engagement, we will incorporate a Chatbot using LLMs, fine-tuning it with OpenAI API or HuggingPhase Models to personalize interactions. Drawing from previous experience, I am particularly interested in this project, having worked on a movie recommendation system using NLP techniques.

Approach for Inventory Forecasting API Development:

Data Collection:

We will ensure accuracy in forecasting by meticulously gathering past inventory data and labeling it for training purposes.

Model Selection:

To achieve precise predictions, we will begin with classical Regression models like Linear Regression, Random Forest Regression, or XGBoost Regression, evaluating their performance against established metrics.

Deep Learning Techniques:

In pursuit of greater accuracy, we will explore deep learning techniques, employing Artificial Neural Networks (ANNs) using TensorFlow or PyTorch to refine our forecasts.

Training and Integration:

Following model selection, we will train our chosen models with past inventory data, seamlessly integrating the forecasting API into our system.

Similar Project Approach:

Leveraging insights gained from past projects, such as the second-hand car price prediction endeavor, we will employ regression models and refine them using ANNs for enhanced efficiency.

Credentials and Experience:

ML/DL Experience:

Extensive experience with ML and DL techniques. Currently learning Generative AI to expand skill set.

Educational Background:

Pre-final Year student at NIT, Jalandhar. Worked on research projects with college faculty.

Achievements:

Participated in Google Solution Challenge with team ranking among top 100 worldwide.

I am quite excited in building this project and looking forward to the opportunity to collaborate.

[aman21376]

Weekly Learnings & Updates (To be filled by contributor)

Week1:

• Researched various aspects of the project, focusing on enhancing search functionality using AI/ML/API models.
• Investigated potential algorithms that could be used for improving search results.
• Analyzed the strengths and weaknesses of different search algorithms.
• Searched for open-source resources and tools that could be integrated into the project.
• Discussed initial ideas and project scope with mentor.

Week2:

• Selected Elasticsearch as the primary search engine for the project after discussing with the mentor.
• Learned about the features and capabilities of Elasticsearch.
• Prepared a low-level design document outlining the system architecture and workflow.
• Created an approach note detailing the steps to be followed for implementation.

Week3:

• Learned how to use Docker Compose for easy deployment and management of Elasticsearch.
• Implemented Elasticsearch on local machines and tested its basic functionality.
• Researched various ML models that can be combined with Elasticsearch to enhance search capabilities.

Week4:

• Started implementing Elasticsearch in different databases such as Amazon, Flipkart, and Myntra products.
• Conducted extensive testing to evaluate the output and performance of Elasticsearch across different databases.
• Indexed the databases inside Elasticsearch, ensuring compatibility with built DB/SQL structures.
• Analyzed search results from various datasets to identify patterns and anomalies.

Week5:

• Achieved functional search capability on the Myntra dataset.
• Implemented advanced search features such as semantic meaning, vectorization, and mean cosine similarity scores.
• Developed a scoring mechanism for each search query and implemented a ranking system based on the scores.
• Presented progress and results to the mentor, incorporating his feedback into the next phase of development.

Week 6:

• Performed spell correction on search queries to improve accuracy.
• Implemented autofill functionality to provide real-time suggestions as users type.
• Added multi-language support to handle search queries in different languages.
• Developed a frontend interface where users can perform searches and view the results.
• Conducted tests to ensure the new features work seamlessly and provide accurate results.

Weekly Learnings & Updates

Week 7:

• Researched various time series forecasting techniques like Exponential Smoothing, ARIMA, and Prophet using Darts.
• Set up the environment with necessary libraries (darts, pandas, numpy), and verified Darts installation by running sample models.
• Loaded and cleaned the dataset (sales data), handled missing values, and converted the date column to the proper datetime format.
• Visualized the time series data, identifying trends, seasonality, and missing data points.

Week 8:

• Implemented Exponential Smoothing using Darts for the Store 1 dataset, forecasting for 5 products.
• Compared the forecasted results from Exponential Smoothing with the actual sales data to evaluate performance.

Week 9:

• Explored and applied ARIMA and Prophet models on the same data to compare results and learn about the model differences.
• Conducted initial tests with various hyperparameters for ARIMA and Prophet to improve model accuracy.

Week 10:

• Explored probabilistic forecasting with models like DeepAR and TCN using Darts.
• Experimented with ensemble models by combining Exponential Smoothing, ARIMA, and Prophet results to improve accuracy.

Week 11:

• Focused on improving forecast accuracy by testing models with more recent data and tweaking settings for Prophet and ARIMA.
• Prepared visualizations to present the forecast results and compared the predicted vs. actual sales values.
• Reviewed the models and results with the mentor for any final feedback and changes.

Week 12:

• Finalized the forecasting models and deployed them for Store 1 and 5 products.
• Worked on creating a visual dashboard to display the forecast results, making it easy to compare predictions with real data.
• wrapped up the time series forecasting project.