Generative AI Integration Services

Looking for World Class PepTalks?

Chatbot

To validate the idea of Peptalk AI, Peptalk developed its MVP version first. The app aimed to transform the way clients find and book experts to organize meetings and events through a smart chatbot interface. The platform provided a seamless experience for users looking to enhance their meetings with specialized insights.

PepTalk is a company dedicated to connecting clients with the right experts for their team meetings and events. We planned to simplify this process with a comprehensive database of experts categorized by various criteria. Find out more about PepTalk at https://www.getapeptalk.com/.

Share
Overview Image

What The Client Required

PepTalk wanted to build a smart chatbot that could engage in natural language conversations with potential clients. The chatbot should also understand users' requirements and suggest suitable experts based on their needs.

  • Collect Keywords: The chatbot prompts users to type keywords or phrases related to their requirements, such as topic, purpose, method, budget, location, duration, industry, and format.  
  • Search For Expert: After accessing the keyword, the chatbot uses text-embedding techniques to analyze user inputs and search for experts in the database. The recommended experts closely match the specified criteria. The chatbot also identifies experts with similar skills, expertise, and experience based on semantic similarity.  
  • View Expert Detail: The chatbot helps users seamlessly access detailed information about experts. Users can request specific data or receive default insights, such as the expert's full biography and available session topics they can address. Plus, the chatbot paraphrases content naturally to ensure users can understand. 
  • Submit Request Information: After users choose a suitable expert, the chatbot streamlines the booking process by gathering essential details like email, phone number, and specific requirements. Then, it submits the request on behalf of the users. The booking workflow starts here.  

Our Technologies

Saigon Technology employed the following technologies to run the project: 

  • Python  
  • Langchain  
  • ChromaDB  
  • FastAPI  
  • Jinja2  
  • Celery  
  • PostgreSQL  
  • Redis  
  • OpenAI  
  • Websocket  
  • Docker  
  • Angular  

Challenges

  • The gap between AI expertise and client domain knowledge made it difficult to define the chatbot's expected outcomes accurately.  
  • The chatbot operated on a third-party service, and we couldn't get direct insights into the logic employed by this service to generate results.  
  • There was a trade-off between performance and cost when using OpenAI. Hence, the project needed a balanced approach to minimize costs while ensuring the chatbot still worked accurately.  
  • The evolving techniques and frameworks required us to update the code constantly.  
Challenges Image
Solutions Image

Solutions

  • Defined clear and measurable acceptance criteria for the upcoming software update to ensure it aligned with user needs and performance benchmarks before deployment.  
  • Identified and addressed any potential issues early in the process.  
  • Finalized tasks and scopes before development began as we followed the fixed-price model.  
  • Used both large and small OpenAI models to balance speed and accuracy in chatbot responses.  
  • Continuously monitored framework repositories for big or latest updates and decided whether to update, refactor, or modify the existing code.  

Reporting

Saigon Technology

We had internal daily meetings to improve workflows and make daily work reports easier. At each milestone, our team and the client gathered in review sessions to check progress and collect feedback. We carefully tracked and updated the implementation status based on client input. Our team documented both in-scope and out-of-scope feedback. Besides, we consistently kept the client updated with the progress of implementation. With regular discussions, the client could understand out-of-scope feedback.  

Building A Team

After assessing the nature of the project and the client's needs, we chose a talented team with skill levels at senior to ensure the quality of the project output. Our team includes: 

Feature Image

Related Projects

AI and ML
Analyze The Object On Image Using Computer Vision Technologies

Analyze The Object On Image Using Computer Vision Technologies

With the help of Computer Vision techniques, this effort seeks to create a system that can evaluate the Object on Image.
Detail arrow
HR
CVParser Documents

CVParser Documents

This project aims at developing an end-to-end system, CV Parser System, to extract important information from a .pdf CV file automatically. As Artificial Intelligence has gained a reputation recently, applying the Computer Vision or Natural Language Processing technologies.
Detail arrow
Natural Language Processing
Natural Language Processing Toolkit

Natural Language Processing Toolkit

The Natural Language Processing Toolkit (NLTK) is a Python-based software application that offers a suite of tools for the purpose of processing natural language data.
Detail arrow
Media and Entertainment
Music Recommendation System

Music Recommendation System

Music Recommendation Systems have become immensely popular, enabling users to explore new songs and artists based on their listening habits and preferences. 
Detail arrow
Computer Vision
Product Recognition

Product Recognition

Utilizing AI-based Computer Vision techniques, the Product Recognition system autonomously detects and categorizes products present within images or videos. Through a comprehensive analysis of the visual attributes of products, including their shapes, colors, and textures
Detail arrow
AI and ML
Skin Analyzing System

Skin Analyzing System

A skin analysis system evaluates skin health and appearance using imaging, machine learning, and data analysis to assess conditions like acne, wrinkles, and sun damage.
Detail arrow
Travel
Semantic Search For Travel Place Document

Semantic Search For Travel Place Document

This project aims at developing a Semantic Searching Engine that could search with meaning not only to find keywords but to determine the intent and contextual meaning of the input sentence. 
Detail arrow
Business
Optical Character Recognition Document

Optical Character Recognition Document

Every day, a vast quantity of textual information is written or printed on tangible paper, such as study-related messages, invoices, periodicals, books, ads, and so on. Paper contamination is a major issue in the corporate world and has obvious environmental consequences.
Detail arrow
Beauty & Wellness
Realitiverse’s Fitness Tracker Platform

Realitiverse’s Fitness Tracker Platform

Realitiverse's Fitness Tracker Platform is designed for Singapore's software market. The platform focuses on fitness lessons and mental health.
Detail arrow

Do you need specific Case studies for Your industry?

We are always ready to assist you. Reach out to us today.
Contact Message Box
Back2Top