Revolutionizing Port Operations Support with AI-Powered Chatbot Solution

INTECH developed an advanced AI chatbot solution for a leading global logistics company. The  solution leverages cutting-edge Large Language Models (LLMs) that dramatically reduces query resolution time and improves operational efficiency.

The client, a major player in global logistics and operating over 80 ports and terminals in more than 40 countries. They use ‘Zodiac’ – a port logistics solution that supports end to end port operations. 

The client faced significant operational challenges in their support system. When a port user encounters issues with the Zodiac system, they contact the support team which then coordinates with the development team to resolve the issue. This process was slow and time-consuming, causing delays in port operations.

Their key challenges included: 

1. Response Delays:

Time-consuming manual support processes resulted in significant waiting times for issue resolution.

2. Resource constraints:

Support staff spent excessive time handling repetitive issues, limiting their ability to address complex problems.

3. Knowledge Management:

Inefficient retrieval and utilization of historical solution data hampered quick resolution of recurring queries.

4. Consistency Concerns:

Inconsistent response quality negatively impacted operational efficiency and user satisfaction.

INTECH solved these challenges by implementing a sophisticated dual-LLM chatbot system, equipped with state-of-the-art AI technology.

1. Data Preparation:

The system begins by uploading historical data of resolved issues into the Chatbot for training purposes. 

This data is then extracted and broken down into manageable segments for efficient processing, where each text segment is converted into vector representations (Embeddings) for better processing.

2. Core System Implementation:

The system uses two distinct Large Language Models – GPT3.5 and Llama2. 

When a user poses a question, the FAISS engine conducts a comparison against the stored text chunks to identify segments with the highest semantic similarity. 

These selected segments are then processed by the chosen language model to generate appropriate responses based on the relevant information from the uploaded files.

3. System Refinement:

The Chatbot is designed to improve over time through two methods: continuous feeding of relevant data and human feedback mechanisms. 

This allows the model to be refined incrementally, ensuring better accuracy and more relevant responses as the system matures.