# US 2025/0165859 A1 · Auto-tracking with just-in-time training and goal seeking ai agents

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2025/0165859 A1
- Document ID: US20250165859A1
- Status: Published application
- Kind: A1
- Application number: 18/822,206
- Filed: 2024-09-01
- Published: 2025-05-22
- Earliest claimed: 2022-10-10
- Inventors: ALBERTO VIDAL; MANU RAO; PRASAD PANCHALAN
- Web page: https://patentgrep.com/patent/US20250165859A1

## CPC

- G06N 20/00: Machine learning
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

In one aspect, a system for auto-tracking with just-in-time training and goal seeking AI agents comprising: a plurality of asset trackers, wherein each asset tracker tracks one or more IoT assets and obtains a set of IoT data; one or more communications hubs; a base station; one or more communication networks; wherein each asset tracker is configured to be in communication with a base station and one or more of the communications hubs; wherein in the one or more communications hubs are configured to be in communication with one or more of the mobile units, and the one or more network; wherein the base station is configured to be in communication with the plurality of asset trackers; a server computing device configured to be in communication with the one or more networks, wherein the server computing device is further configured to implement the following logic: wherein one or more asset trackers transmit raw data directly to the server computing device, instead of device ML model-derived inference or device computed data, storing the raw data of the one or more asset trackers in the server computing device without analytical computation; receiving a user query; initiating a just-in-time process to find an answer to the user query; communicating the question to an AI agent in the server computing device; with the AI agent: seeking a goal of answering the question, by breaking the question down into multiple steps, and executing the multiple steps until the goal is reached.

## Claims

### Claim 1

A system for auto-tracking with just-in-time training and goal seeking AI agents comprising:
a plurality of asset trackers, wherein each asset tracker tracks one or more IoT assets and obtains a set of IoT data;
one or more communications hubs;
a base station;
one or more communication networks;
wherein each asset tracker is configured to be in communication with a base station and one or more of the communications hubs;
wherein in the one or more communications hubs are configured to be in communication with one or more of the mobile units, and the one or more network;
wherein the base station is configured to be in communication with the plurality of asset trackers;
a server computing device configured to be in communication with the one or more networks, wherein the server computing device is further configured to implement the following logic:
wherein one or more asset trackers transmit raw data directly to the server computing device,
instead of device ML model-derived inference or device computed data, storing the raw data of the one or more asset trackers in the server computing device without analytical computation;
receiving a user query;
initiating a just-in-time process to find an answer to the user query;
communicating the question to an AI agent in the server computing device;
with the AI agent:
seeking a goal of answering the question, by breaking the question down into multiple steps, and
executing the multiple steps until the goal is reached.

### Claim 2 (depends on 1)

The computerized system of claim 1, wherein the AI agent uses a plurality of internal and external services to answer the question including a specific just-in-time training of the raw tracking data with a specified ML training methodology.

### Claim 3 (depends on 2)

The computerized system of claim 2, wherein specific ML training methodology and ML algorithm are selected just-in-time for the AI agent to answer the user query with a higher precision.

### Claim 4 (depends on 3)

The computerized system of claim 3, wherein the AI agent runs the raw tracking data through multiple pre-trained models and selects an answer that best fits the user query.

### Claim 5

A system for auto-tracking with just-in-time training and goal seeking ai agents comprising:
a plurality of asset trackers, wherein each asset tracker tracks one or more IoT assets and obtains a set of IoT data;
one or more communications hubs;
a base station;
one or more communication networks;
wherein each asset tracker is configured to be in communication with a base station and one or more of the communications hubs;
wherein in the one or more communications hubs are configured to be in communication with one or more of the mobile units, and the one or more network;
wherein the base station is configured to be in communication with the plurality of asset trackers;
a server computing device configured to be in communication with the one or more networks, wherein the server computing device is further configured to implement the following logic:
asset tracker provides data to a cloud-based server at preset intervals;
storing the data for future use;
based on a user query, with a tracking system of the server computing device accessing the data and applying the data to a machine learning mode; and
creating and training a machine-learning (ML) model on the fly based on the query to provide an insight to the user related to the user query.
