Welcome to your weekly AI Newsletter from AITechCircle!

I’m building and implementing AI solutions and sharing everything I learn along the way…

This week was loaded with many new LLM releases and breakthroughs in image Generation. Before starting this week’s newsletter, let me say Eid Mubarak to whoever is celebrating, and I believe this will bring you peace and prosperity. Eid Mubarak in Ghibli-effect 🙂

Check out the updates from this week! Please take a moment to share them with a friend or colleague who might benefit from these valuable insights!

Today at a Glance:

• Increasing the adoption and effective use of AI and Generative AI technologies in various use cases or business processes
• Generative AI Use cases repository
• AI Weekly news and updates covering newly released LLMs
• Courses and events to attend

Agentic AI – Exploring Oracle’s Generative AI Agents

Early this morning, I got a chance to explore the newly released ​Oracle Generative AI Agent platform​. Below are some of the early setups I was playing around with.

Before you get into the nitty-gritty of the platform, a few thoughts: The best part is that it is easy to develop or build, offers a pleasant user experience, and covers wider data sources.

​

The OCI Generative AI Agents platform is an enterprise-ready service that empowers businesses to quickly create, deploy, and oversee AI agents. With it, you can develop virtual assistants/chatbots that deliver personalized, relevant, and engaging interactions. There are a few features to highlight

​

• Pre-built tools: comes with ready-to-use tools for agents to address everyday organization needs:
  • RAG (Retrieval Augmented Generation): chat with unstructured enterprise data using natural language.
  • SQL: chat with structured enterprise data using natural language.
• Custom tools support: extends agent’s functionality with custom tools:
  • Function calling: create callable functions – the agent identifies which function to invoke and passes appropriate parameters.
  • API endpoint calling (limited availability): define tools that interact with external APIs – the agent executes the tool and returns the final result (e.g., content generation, data retrieval, or system interaction).

​

Let’s start…

Login into your Oracle Tenancy and go to Analytics & Tools > Generative AI Agents

Select Create Agent

Here, you need to enter basic details like name, welcome message, etc.

After following, you will see this below page and select Create tool.

​

Now you have to select the prebuilt tool, what you want to use,

1. RAG
2. SQL
3. Custom Tool

RAG: The pre-built RAG (Retrieval-Augmented Generation) tool, part of the Agent’s platform, has enhanced how users retrieve and generate information from vast knowledge sources.

SQL: Introducing the pre-built NL2SQL tool as part of the Agent’s platform experience. This tool changes how users interact with structured organization data by enabling data retrieval and analysis through natural language.

Custom Tool: It extends the agent’s functionality with custom tools

Select RAG as you are going to create an HR Policy agent

Also, create a knowledge base for the documents you will pass on to this RAG agent. These are the policy documents in PDF format.

You have three options for the Knowledge base:

1. Object Storage – OCI Object storage
2. OCI OpenSearch – You must have documents chunked to files with less than 512 tokens each, and you must have ingested and indexed those documents in OpenSearch before you continue
3. Oracle AI Vector Search – This option is for data in Oracle Database 23ai.

​

​

After the tools setup is done, you now need to move to create a Setup Agent Endpoint

Here, you can see different options for:

• Human in the loop
• and Guardrails
• Content moderation

Prompt (PI) injections

Personal Identifiable Information (PII) protection

Move next to the setup.

You will accept the Llama 3 license agreement and submit

you can see your agent is being created

Once this creation process is complete, you can access the chat function and chat with your Agent.

If you need to have SQL as a Tool in the Agent

I will leave this to next week to cover; we will create the sample data on an autonomous database and use this SQL agent.

​

Call for Action:

Try it out yourself. Enter your OCI Tenancy and head to the OCI Generative AI Agent platform. Share your experience of exploring this.

​

Weekly News & Updates…

Last week’s AI breakthroughs marked another leap forward in the tech revolution.

1. ​ARC-AGI-2 It’s an AI benchmark designed to measure general fluid intelligence, not memorized skills – a set of never-seen-before tasks that humans find easy but current AI struggles with. It keeps the same format as ARC-AGI-1 while significantly increasing the signal strength it provides about a system’s actual fluid intelligence
2. ​Qwen2.5-VL-32B-Instruct optimizes the VLM with reinforcement learning, and we have found significant improvements in human preference and also mathematical reasoning
3. ​Mistral Small 3.1: this new model comes with improved text performance, multimodal understanding, and an expanded context window of up to 128k tokens
4. LG AI Research has released the EXAONE Deep LLM model, specializing in math, science, and coding tasks. EXAONE Deep pushes the boundaries of AI’s role in professional fields and everyday life
5. OpenAI releases speech-to-text and text-to-speech audio models in the API—making it possible to build more robust, customizable, and intelligent voice agents
6. ​DeepSeek-V3-0324, with a significant boost in reasoning performance and Stronger front-end development skills.
7. ​Gemini 2.5 Pro Experimental is a thinking model designed to tackle increasingly complex problems

The Cloud: the backbone of the AI revolution

• NVIDIA NIM Microservices Now Available to Streamline Agentic Workflows on RTX AI PCs and Workstations link​
• Oracle announced New AI Infrastructure Capabilities with NVIDIA Blackwell for Public, On-Premises, and Service Provider Clouds link​

Generative AI Use Case of the Week:

Several Generative AI use cases are documented, and you can access the library of generative AI Use cases. Link​

Real-Time AI-Based Video Content Moderation

Implementing AI and Generative AI models to automatically detect and censor inappropriate visual and audio elements (nudity, explicit language, violence) in real-time broadcasts and streaming video platforms.

Challenge:

Manual moderation of live video streams is resource-intensive, error-prone, and difficult to scale effectively due to high latency and large data volumes.

Solution:

• Utilize real-time computer vision models (e.g., YOLOv8, EfficientDet, or Meta’s Segment Anything Model [SAM], LLMs) to identify inappropriate visuals.
• Deploy audio processing algorithms (e.g., OpenAI’s Whisper or Google’s Speech-to-Text API coupled with NLP models like GPT-4) to recognize and censor inappropriate speech.
• Use generative models to automatically censor problematic content by visually replacing it seamlessly with appropriate alternatives.

Benefits:

• Improved accuracy and reduced human moderator workload.
• Instantaneous censorship for real-time compliance.
• Reduction of regulatory fines due to content violations.

Impact:

• Enhanced reputation through compliance and improved viewer trust.
• Reduction of operating costs due to automation.
• Improved scalability for high-volume video content distribution.

Favorite Tip Of The Week:

Here’s my favorite resource for the week.

This is the best resource I have ever known about setting up personal digital security, and I plan to implement it in a week.

Andrej Karpathy @karpathy I wrote a quick new post on “Digital Hygiene”. Basically there are some no-brainer decisions you can make in your life to dramatically improve the privacy and security of your computing and this post goes over some of them. Blog post link in the reply, but copy pasting below 9:14 PM • Mar 18, 2025 3603 Retweets 27097 Likes Read 737 replies

​

Potential of AI:

Implementing GPT From Scratch

​

Things to Know…

​The 2025 Edge AI Technology Report provides a comprehensive analysis of the advancements and applications of edge

AI technology across various industries. It highlights the shift from centralized cloud AI to edge-centric AI, driven by the need for lower latency and enhanced data privacy.

The report explores significant technological breakthroughs that enable edge AI, including specialized chips and real-time algorithms, and emphasizes the crucial role of ecosystem partnerships in propelling edge AI forward. It also outlines future directions for edge AI, suggesting its increasing integration into everyday devices and systems to improve operational efficiency and innovation

The Opportunity…

Podcast:

• This week’s Open Tech Talks episode 156 is “Building and Growing an AI Startup with Sébastien Night” he is a Co-Founder of OneTake AI.

Apple | Amazon Music​

Building and Growing an AI S… Mar 23 · OPEN Tech Talks: Technol… 43:20   ​

Courses to attend:

• ​Vibe Coding 101 with Replit from DeepLearning AI

Events:

• ​Dubai AI Festival 2025, April 23-24, 2025, Dubai, UAE

Tech and Tools…

• ​Agno is a lightweight library for building Multimodal Agents. It exposes LLMs as a unified API and gives them superpowers like memory, knowledge, tools and reasoning
• ​GenAI Agents: Comprehensive Repository for Development and Implementation

The Investment in AI…

• ​Frontera Health, an AI-enabled autism care company, launched in mid-February with $32 million in seed funding
• ​Camb.ai: The artificial intelligence-powered speech-to-speech translation startup has a new multi-year partnership with LFP, the governing body of French soccer; and, separately, closed an $11 million Pre-Series A funding round, bringing its total funding to $15.5 million.

And that’s a wrap!

Thank you, as always, for taking the time to read.

I would love to hear your thoughts. Please reply and share what you found most valuable this week. Your feedback means a lot to me.

Until next week,

Kashif Manzoor