---
title: 'Building an AI System from a Job Interview Use Case'
url: 'https://mcavdar.com/blog/building-an-ai-system-from-a-job-interview-use-case'
markdown: 'https://mcavdar.com/blog/building-an-ai-system-from-a-job-interview-use-case.md'
date: '2025-07-16'
description: 'Recently, I passed a job interview where we discussed a practical use case related to intelligent document processing and automation. In this blog series, I’ll attempt to implement key parts of that use case to evaluate how well I can design such a system and integrate the relevant tools and softwa…'
taxonomy:
  tag:
    - langchain
    - zoho
    - data-ingestion
    - llamaindex
    - rag
    - OAuth2
    - curl
    - Agentic-AI
    - AI-Engineering
---

# Building an AI System from a Job Interview Use Case

## [Building an AI System from a Job Interview Use Case](https://mcavdar.com/blog/building-an-ai-system-from-a-job-interview-use-case)

    16th Jul 2025   [langchain](https://mcavdar.com/tag:langchain#body-wrapper) [zoho](https://mcavdar.com/tag:zoho#body-wrapper) [data-ingestion](https://mcavdar.com/tag:data-ingestion#body-wrapper) [llamaindex](https://mcavdar.com/tag:llamaindex#body-wrapper) [rag](https://mcavdar.com/tag:rag#body-wrapper) [OAuth2](https://mcavdar.com/tag:OAuth2#body-wrapper) [curl](https://mcavdar.com/tag:curl#body-wrapper) [Agentic-AI](https://mcavdar.com/tag:Agentic-AI#body-wrapper) [AI-Engineering](https://mcavdar.com/tag:AI-Engineering#body-wrapper)  

Recently, I passed a job interview where we discussed a practical use case related to intelligent document processing and automation. In this blog series, I’ll attempt to implement key parts of that use case to evaluate how well I can design such a system and integrate the relevant tools and software.

#### The Use Case

We are given the following scenario:

- We use Zoho CRM/Email to ingest data — mostly in PDF format.
- We need to parse the contents of these documents.
- Finally, we aim to build an agentic AI system using RAG (Retrieval-Augmented Generation) to allow users to query the documents and automate Business Intelligence (BI) tasks.

#### Roadmap

Part 1 – Ingest Data (Zoho API)

Part 2 – Parse Data (LlamaIndex)

Part 3 – Build AI Agent (LangChain + RAG)

---

### Part 1 – Ingest Data via Zoho API

To begin, I set up a custom domain to test the Zoho integration. Authentication with the Zoho API is handled via OAuth 2.0.

#### Step 1: Create OAuth Credentials

I followed the guide at [Zoho OAuth Request Docs](https://www.zoho.com/writer/help/api/v1/oauth-requests.html) and first created a [Server-based Application](https://api-console.zoho.eu), which provided me with a **Client ID** and **Client Secret**.

To request the authorization code:

```
https://accounts.zoho.eu/oauth/v2/auth?
scope=ZohoMail.messages.READ,ZohoMail.accounts.READ&
client_id=1000.xxxx&
response_type=code&
access_type=offline&
redirect_uri=https://yourdomain/oauth/callback
```

This redirects to:

```
https://yourdomain/oauth/callback?
code=1000.yyyy.zzzz&
location=eu&
accounts-server=https%3A%2F%2Faccounts.zoho.eu
```

#### Step 2: Get Access Token

Run this:

```
curl --request POST \
  --url "https://accounts.zoho.eu/oauth/v2/token" \
  --data "grant_type=authorization_code" \
  --data "client_id=1000.xxxx" \
  --data "client_secret=your_client_secret" \
  --data "redirect_uri=https://yourdomain/oauth/callback" \
  --data "code=1000.yyyy.zzzz"
```

> Note: Don’t forget to replace your\_client\_secret with your actual client secret.

Sample Response:

```
{
  "access_token": "1000.ttt",
  "refresh_token": "1000.uuu",
  "scope": "ZohoMail.messages.READ ZohoMail.accounts.READ",
  "api_domain": "https://www.zohoapis.eu",
  "token_type": "Bearer",
  "expires_in": 3600
}
```

#### Step 3: Get Account ID

You need your Zoho Mail account ID to access your messages/mails:

```
curl --request GET \
  --url https://mail.zoho.eu/api/accounts \
  --header "Authorization: Zoho-oauthtoken 1000.ttt"
```

Note: This may return an error if your token scope doesn't include ZohoMail.accounts.READ.

#### Step 4: Fetch Emails

Once you have the account ID:

```
curl --request GET \
  --url "https://mail.zoho.eu/api/accounts/youraccountid/messages/view?limit=10" \
  --header "Authorization: Zoho-oauthtoken 1000.ttt"
```

> Note: Don’t forget to replace youraccountid with your actual account id.

Example Response:

```
{
  "status": {
    "code": 200,
    "description": "success"
  },
  "data": [
    {
      "summary": "Hey MC, Welcome 🙏 Mahmut ÇAVDAR...",
      "subject": "Re: Test Email",
      "fromAddress": "mahmutcvdr@gmail.com",
      "hasAttachment": "0",
      "folderId":"7192274xxxx000002014",
      "messageId":"17525962xx222005600",
      ....
    }
  ]
}
```

If there's no attachment, "hasAttachment": "0". If there is an attachment, you can fetch its metadata:

#### Step 5: Get Attachment Info

Once you have the account ID:

```
curl "https://mail.zoho.eu/api/accounts/youraccountid/folders/7192274xxxx000002014/messages/17525962xx222005600/attachmentinfo?includeInline=true" \
-X GET \
-H "Accept: application/json" \
-H "Content-Type: application/json" \
-H "Authorization:Zoho-oauthtoken 1000.ttt"
```

Example Response:

```
{
  "status": { "code": 200, "description": "success" },
  "data": {
    "attachments": [
      {
        "attachmentName": "CV_Cavdar.pdf",
        "attachmentId": "139719xxx010000"
      }
    ]
  }
}
```

#### Step 6: Download Attachment

Once you have the account ID:

```
curl "https://mail.zoho.eu/api/accounts/youraccountid/folders/7192274xxxx000002014/messages/17525962xx222005600/attachments/139719xxx010000" \
-X GET \
-H "Accept: application/octet-stream" \
-H "Content-Type: application/json" \
-H "Authorization:Zoho-oauthtoken 1000.ttt" 
-o cv.pdf
```

And voilà — you have the attachments!

I even found an issue in the Zoho API documentation. The example on [this page](https://www.zoho.com/mail/help/api/get-attachment-content.html) is incorrect — it shows the example for "Get Email Attachment Info" instead of "Get Email Attachment Content".

Wrong example:

```
curl "https://mail.zoho.com/api/accounts/12345678/folders/9000000002014/messages/1710915488416100001/attachmentinfo" \
-X GET \
-H "Accept: application/json" \
-H "Content-Type: application/json" \
-H "Authorization:Zoho-oauthtoken *****"
```

Correct Example:

```
curl "https://mail.zoho.com/api/accounts/12345678/folders/9000000002014/messages/1710915488416100001/attachments/139712110853010000" \
-X GET \
-H "Accept: application/octet-stream" \
-H "Content-Type: application/json" \
-H "Authorization:Zoho-oauthtoken *****"
```

The bug has been reported [here](https://help.zoho.com/portal/en/community/topic/bug-incorrect-example-in-the-documentation).

---

### Next Steps

In Part 2, I’ll use [LlamaIndex](https://www.llamaindex.ai) to parse PDF files and structure their contents for downstream processing. I’ll likely use this [sample invoice dataset](https://github.com/femstac/Sample-Pdf-invoices/tree/main) as a starting point.

In Part 3, I’ll bring in [LangChain](https://www.langchain.com/) to build a agentic system that can run user queries over the ingested content and support basic BI automation tasks.

 [ Previous Post](https://mcavdar.com/blog/crawl4ai-in-action-real-world-use-cases-for-smarter-web-scraping) [Next Post ](https://mcavdar.com/blog/the-easiest-way-to-get-started-with-llamaindex-create-llama)

#### Syndicate

 [ Atom 1.0](https://mcavdar.com/blog.atom) [ RSS](https://mcavdar.com/blog.rss)

---

## Navigation

- Parent: [Blog](https://mcavdar.com/index.md)
- Previous: [The easiest way to get started with LlamaIndex: create-llama](https://mcavdar.com/blog/the-easiest-way-to-get-started-with-llamaindex-create-llama.md)
- Next: [Crawl4AI in Action: Real-World Use Cases for Smarter Web Scraping](https://mcavdar.com/blog/crawl4ai-in-action-real-world-use-cases-for-smarter-web-scraping.md)