Governing Agentic AI: How Organizations Can Balance Autonomy, Risk, and Accountability

1. Advanced Analytics in HR HR teams always struggle with mammoth amounts of data which can be transformed into meaningful insights with the help of Advanced Analytics. The applications of Advanced Analytics can spearhead the domains of recruitment, learning and development of the employees and help manage their performance. It can aid HR to identify the shortcomings in the existing resources and help decide to improve the skills of existing employees or hire competent new talent. The added advantage of HR analytics is that it is efficient in analysing employee performance by categorising them according to their individual calibre. 2. Advanced Analytics in Banking Advanced Predictive Analytics has the potential to gear up the banking sector by assisting banks to adopt better strategies assessing and predicting customer behaviour. It can safeguard customer accounts from cyber-attacks and can detect fraudulent activities. Real-time alert messages are the most exclusive of the features which can shield against cyber-attacks to a big extent. 3. Advanced Analytics in Retail The uncertain economy creates challenging scenarios for retailers. But Advanced Analytics functions act as a handy tool to make use of customer insights from the silos of data and provide a personalised and tailored marketing experience. It can be both cost saving and can unleash business growth to a great extent. 4. Advanced Analytics in Manufacturing Manufacturing Analytics harvests revolutionary output from the manufacturing process, supply chain production, and customer delivery. The global manufacturing market is predicted to grow up to $28.5 billion by 2026 and the growth can be multiplied using the insights generated by analytics. Tools for asset management, product development, inventory management, and predictive maintenance can generate insightful options to rake in revenue amidst the limitations of the times. 5. Advanced Analytics in Finance Financial sector produces a huge quantity of structured and unstructured data. The finance consultants or analysts should exploit this data with the help of Advanced Analytics to tackle financial troubles and to identify customer related requirements and issues. Advanced Analytics also helps to implement business plans and eliminate fraudulent activities. Financial institutions like loaning bodies, banks, trading firms, and insurance companies can improve their performance and revenue by deploying Advanced Analytics. 6. The Future of Advanced Analytics Advanced Analytics is evolving and has become the indispensable requirement to run businesses. Due to the worldwide reception and adoption by the business community, analytics and information services sectors are predicted to have the fastest growth rate from 16% to 18% over the next five years starting 2022. As the market is getting competitive, planning and implementing strategic investments with data analytics is crucial to create new services and added revenue. How Beinex can act as a torchbearer Beinex has impacted and accelerated the growth of its customers and clients. With its impeccable offerings, it has partaken to build, market and expand businesses worldwide. The services provided by Beinex are enlisted below bracketed into respective domains: • Predictive Analytics This outstanding service offers you predictive insights to make resolutions which in turn can power your business. • Artificial Intelligence To cut short expenses and to attain target within a short period of time, make use of our AI tools or applications. It is beneficial to aggregate and automate data processing. • Machine Learning A vast quantity of data can be analysed using algorithms which can accelerate the decision-making process. With Machine Learning, specific and relevant data can be extracted and utilised to leverage businesses. • RPA Robotic Process Automation or software robots can accelerate digital transformation by emulating human actions. Automation of workflows, greater degree of accuracy, better productivity, high efficiency, and lesser costs are a few major takeaways from RPA which can accelerate business growth. Conclusion Advanced Analytics is here to stay, as organisations have started to rely on analytics more than ever before to predict business outcomes. It is sure to govern the business world as it has gained trust as a superpower which is accessible to all but chosen by only a few.

Businesses are indulging in the continuous endeavour to leverage the full potential of their data, and they are constantly seeking innovative solutions for the same. A solution capable of gaining traction is the integration of Meta Llama 3 into the Snowflake ecosystem. Beinex has brought this to fruition by establishing an innovative work model that guarantees an application that functions within Snowpark Container Services and harnesses the capabilities of Llama 3. Figure 1: Screenshot from DocAI. Zaki Document Chatbot (DocAI) tapping into Llama 3 by Meta and running in the Snowflake ecosystem. Beinex tested Llama 3 on its in-house DocAI, a solution that runs on Snowflake using Snowpark Container services. The DocAI chatbot solution offers the flexibility to chat with documents such as PPT, PDF, word files, and text files. It currently uses llama3. Llama 3 is a major leap forward, establishing new standards for large language models. Its extensive training data, improved quality, and increased context length make it a powerful choice for document-related tasks, including our DocAI chatbot solution. The article will explain how Beinex deployed Llama 3 in the Snowflake ecosystem in the upcoming sections. Recently, notable advances have been made in large language models — sophisticated natural language processing (NLP) systems equipped with billions of parameters. These models have demonstrated remarkable abilities, including generating creative text, solving complex mathematical theorems, predicting protein structures, and more. They illustrate the immense potential benefits that AI can offer to billions of people on a global scale. Meta’s Llama (Large Language Model Meta AI), a state-of-the-art foundational large language model, is designed to support researchers in advancing their work within AI. By providing access to smaller yet highly efficient models like Llama, Meta aimed to empower researchers who may not have access to extensive infrastructure to delve into the study of these models. This democratization of access is pivotal in fostering innovation and progress in this dynamic and crucial field.

What is Llama 3?

Meta's latest advancement in the LLM (Large Language Model) series is Llama 3, the most sophisticated model with considerable advancements in performance and AI capabilities. Llama 3, built upon the architecture of Llama 2, is offered in 8B and 70B parameters, each featuring a base model and an instruction-tuned version tailored to enhance performance in specific tasks, particularly AI chatbot conversations. According to Meta, Llama 3 sets a new standard for open-source models, rivalling the performance of proprietary models available today. Llama 3 models will soon be accessible across various platforms, including AWS, Google Cloud, Hugging Face, Databricks, Kaggle, IBM Watson, Microsoft Azure, NVIDIA NIM, and Snowflake. Capabilities such as reasoning, code generation, and instruction following have seen substantial enhancements, rendering Llama 3 more adaptable and controllable. Meta plans to introduce additional capabilities, longer context windows, expanded model sizes, and enhanced performance. Utilizing Llama 3 technology, Meta AI emerges as one of the premier AI assistants globally, offering intelligence augmentation and support across various tasks, including learning, productivity, content creation, and connection facilitation.

Llama 2 vs Llama 3

According to Meta, the newly introduced models, Llama 3 8B with 8 billion parameters and Llama 3 70B with 70 billion parameters, represent a significant advancement in performance compared to their predecessors, Llama 2 8B and Llama 2 70B. Meta describes these models as a ‘major leap’ in performance. Llama 2 serves research and commercial purposes, excluding the top consumer companies globally. Llama 2 boasts enhancements such as training on 40% more data, doubling the context length, and leveraging a vast dataset of human preferences to ensure safety and helpfulness, backed by over 1 million annotations. On the other hand, Llama 3 represents the next step in Meta AI's LLM evolution, catering to research and commercial applications, provided monthly active users are under 700 million. Positioned as the successor to Llama 2, Llama 3 showcases state-of-the-art performance on benchmarks and is lauded by Meta as the 'best open-source model of their class.'

Ollama

There were times when accessing Large Language Models (LLMs) was restricted to cloud APIs offered by major providers like OpenAI and Anthropic. While these cloud API providers continue to dominate the market with user-friendly interfaces facilitating easy access for many users, it's important to recognize the trade-offs users make beyond the costs associated with pro plans or API usage. This trade-off involves granting providers full access to chat data. For those seeking to securely run LLMs on their hardware, the alternative has typically involved training their LLMs. Ollama, an open-source application, is designed to enable users to run, create, and share large language models locally through a command-line interface on MacOS and Linux. With Ollama, running LLMs on personal hardware requires minimal setup time. It caters to individuals seeking to run LLMs on their laptops, maintain control over their chat data without involving third-party services, and interact with models through a straightforward command-line interface. Additionally, Ollama offers various community integrations, including user interfaces and plugins for chat platforms.

Deploying Llama 3 in the Snowflake Ecosystem: What Beinex Did?

Deploying Llama 3 in the Snowflake Ecosystem means integrating the advanced language capabilities of Llama 3, the latest version of Meta’s language model, into the Snowflake data platform. It represents a significant breakthrough for organizations seeking to maintain control over their data. It allows users to directly leverage Llama 3's powerful natural language processing capabilities within Snowflake for various tasks such as data analysis, querying, and generating insights. 

Users can benefit from enhanced text understanding and generation

Enables more efficient and accurate data processing, analysis, and decision-making

Empowers users to interact with their data more intuitively and naturally

Drives greater productivity and insights within the Snowflake environment

Traditionally, utilizing modern LLMs at scale has necessitated reliance on proprietary services, often requiring organizations to abandon control over their data to service providers. This compromise between data residency and accessing advanced AI capabilities has been a conflict for many. However, a solution has emerged with Beinex's innovative approach and successful Proof of Concept (PoC). Beinex has devised a model ensuring that applications function seamlessly within Snowpark Container Services while harnessing the power of Llama 3. This integration means organizations can now deploy Llama 3 directly within the Snowflake ecosystem, eliminating the need for reliance on external proprietary services. Figure 2: Zaki Document Chatbot (DocAI) in action!

Deploying Llama 3 in the Snowflake Ecosystem: How Beinex Did it?

Here’s a detailed guide on deploying Llama 3 on Snowflake Container Services: Step 1: Create Necessary Objects -- Run by ACCOUNTADMIN to allow connecting to Hugging Face to download the model -- Stage to store LLM models CREATE STAGE <stagename> IF NOT EXISTS models DIRECTORY = (ENABLE = TRUE) ENCRYPTION = (TYPE='SNOWFLAKE_SSE'); -- Stage to store YAML specs CREATE STAGE <stagename> IF NOT EXISTS specs DIRECTORY = (ENABLE = TRUE) ENCRYPTION = (TYPE='SNOWFLAKE_SSE'); <br. -- Image repository CREATE OR REPLACE IMAGE REPOSITORY images; -- Compute pool to run containers CREATE COMPUTE POOL GPU_NV_S MIN_NODES = 1 MAX_NODES = 1 INSTANCE_FAMILY = GPU_NV_S; Step 2: Docker Image Code - ollama FROM ollama/ollama RUN $(ollama serve > output.log 2>&1 &) && sleep 10 && ollama pull llama3 && pkill ollama && rm output.log ENTRYPOINT ["ollama"] CMD ["serve"] Step 3: Tag and Push the Docker Image docker tag ollama .registry.snowflakecomputing.com/db/schema/image respository /ollama docker push .registry.snowflakecomputing.com db/schema/image repository /ollama Step 4: Docker Image - UDF FROM python:3.11 WORKDIR /app ADD ./requirements.txt /app/ RUN pip install --no-cache-dir -r requirements.txt ADD ./ /app EXPOSE 5000 ENV FLASK_APP=app CMD ["flask", "run", "--host=0.0.0.0"] App.py content is given below : from flask import Flask, request, Response, jsonify import logging import re import os from openai import OpenAI client = OpenAI( base_url='http://ollama:11434/v1', api_key="EMPTY", ) model = "llama3" app = Flask(__name__) app.logger.setLevel(logging.ERROR) def extract_json_from_string(s): logging.info(f"Extracting JSON from string: {s}") # Use a regular expression to find a JSON-like string matches = re.findall(r"\{[^{}]*\}", s) if matches: # Return the first match (assuming there's only one JSON object embedded) return matches[0] # Return the original string if no JSON object is found return s @app.route("/", methods=["POST"]) def udf(): try: request_data: dict = request.get_json(force=True) # type: ignore return_data = [] for index, col1 in request_data["data"]: completion = client.chat.completions.create( model=model, messages=[ { "role": "system", "content": "You are a bot to help extract data and should give professional responses", }, {"role": "user", "content": col1}, ], ) return_data.append( [index, extract_json_from_string(completion.choices[0].message.content)] ) return jsonify({"data": return_data}) except Exception as e: app.logger.exception(e) return jsonify(str(e)), 500 Step 6: YAML File spec: containers: - name: ollama image: <SNOW_ORG-SNOW_ACCOUNT>.registry.snowflakecomputing.com/ db/schema/image respository /llama3 resources: requests: nvidia.com/gpu: 1 limits: nvidia.com/gpu: 1 env: NUM_GPU: 1 MAX_GPU_MEMORY: 24Gib volumeMounts: - name: llm-workspace mountPath: /<stage name> - name: udf image: .registry.snowflakecomputing.com/ db/schema/image respository /ollama_udf endpoints: - name: chat port: 5000 public: false - name: llm port: 11434 public: false volumes: - name: llm-workspace source: "@<llm stage_name>" Step 7: Upload YAML File and Create Service Upload the YAML file to the created stage, where the stage name in the YAML file should match the stage created in Step 2. -- Create service create service llama3 IN COMPUTE POOL<name of compute pool created> FROM @dash_stage SPECIFICATION_FILE = '<name of yaml file uploaded>'; Step 8: Create Service Function Create a service function on the service (after it starts). create or replace function llama3(prompt text) returns text service=llama3 endpoint=chat; Check Service Status Use the following command to check the status of the service: SELECT v.value:containerName::varchar container_name, v.value:status::varchar status, v.value:message::varchar message FROM ( SELECT parse_json(system$get_service_status('<service name>')) ) t, LATERAL FLATTEN(input => t.$1) v;

Benefits of Deploying Llama 3 in the Snowflake Ecosystem

Superior AI apps courtesy of Llama 3: Allows organizations to harness the full potential of this advanced LLM.

Llama 3 offers avant-garde capabilities in natural language processing and generation, enabling the development of superior AI applications with enhanced performance, accuracy, and efficiency.

Data residency: Ensures that organizations maintain control over their data residency. With Snowflake's robust data governance and compliance features and Llama 3's deployment within the same ecosystem, organizations can confidently adhere to data residency requirements and regulations, ensuring that sensitive data remains within specified jurisdictions or regions.

Scalability: Provides organizations with remarkable scalability for their AI applications. This seamless scalability empowers organizations to handle large volumes of data and accommodate growing user demands without compromising performance.

Easy monitoring: Simplifies the monitoring and management of AI applications. Snowflake's intuitive interface and comprehensive monitoring tools enable organizations to easily track the performance, usage, and health of their deployed Llama 3 models. It facilitates proactive identification of issues, optimization of resources, and overall better management of AI workloads, ensuring smooth and efficient operations.

By leveraging Llama 3 within Snowflake, organizations can alleviate data sovereignty and control concerns. They can rest assured that their precious and sensitive data remains within their realm of control while still being able to build and scale superior applications. This development marks a significant milestone in bridging the gap between data residency requirements and accessing the most effective LLM technology.

-Mebin Joy

Benefits of AWS Cost Optimization

Optimizing AWS costs is a vital part of effectively utilizing AWS services. Businesses can leverage the wide array of AWS services and pricing models to learn how to optimize costs and save considerable amounts of money. AWS Cost Optimization offers several key benefits to organizations looking to manage and reduce cloud expenses while maximizing efficiency. Let's take a look at the benefits of cost optimization with AWS: Enhanced flexibility: AWS offers flexible and cost-effective purchase plans like Reserved Instances or Amazon EC2 Spot Instances, enabling businesses to select the suitable pricing models that best fit their requirements. Boosted efficiency in resource utilization: Allocate resources effectively by rightsizing your instances, identifying the optimal size for your compute workloads, selecting the right data storage and transfer options, and eliminating unused resources. This prevents overprovisioning and ensures optimal usage of available services. Increased cost savings: AWS identifies unused resources and helps businesses scale down their infrastructure, reducing cloud costs and saving big. AWS cost optimization also ensures that you pay only for what you need, enhancing the efficiency of your cloud operations. Improved budgeting and forecasting: AWS Cost Optimization gives businesses greater visibility into spending patterns, enabling them to forecast costs and create accurate and realistic budgets based on seasonal trends, historical usage patterns, and business growth projections. Increased scalability: By employing cost optimization strategies, businesses can seamlessly scale resources on demand and pay for the requirement only when needed. Optimizing costs for seasonal fluctuations, changing workloads, and business growth ensures scalability in your AWS environment.

A Quick Look into AWS Cloud Cost Optimization Tools

AWS Cost Explorer: A tool that offers insights into the cost and usage of AWS services by giving a detailed picture of AWS spending. Its benefits include: • Identifying areas where costs are optimized • Forecasting future costs • Planning and budgeting AWS costs AWS Budgets: A tool that allows users to create custom cost and usage budgets for AWS services. It enables users to take corrective action by sending alerts when cost and usage thresholds exceed. Its benefits include: • Handling AWS spend by setting and monitoring budgets • Avoiding unforeseen expenses and cost overruns • Providing suggestions for cost optimization AWS Trusted Advisor: A tool offering tailored recommendations for optimizing AWS costs, performance, and security. It is free of charge for all AWS users. Its benefits include: • Delivering actionable insights from usage data • Offering personalized recommendations from usage data • Identifying areas where costs can be lowered AWS Pricing Calculator: A tool that helps users get estimates before choosing a service and enables the assessment of calculations based on the estimates. Its benefits include: • Identifying possible areas for cost-saving • Planning budget effectively • Providing insights into the costs you expect for a service

Optimizing Cloud Costs with AWS Cost Explorer

A robust tool that enables users to visualize, analyze, and monitor AWS costs, AWS Cost Explorer offers a granular view of your AWS spending and usage. It allows you to track your spending over time, detect cost trends, and investigate specific resources to understand where the money is spent. Cost Explorer provides a detailed report on cost, usage, and Reserved Instances and a main graph that helps you drill down into costs and usage. Users can access data for up to the last thirteen months, get suggestions for RI purchases, and predict expenses for the next twelve months. Besides, Cost Explorer offers insights into user spending by identifying areas for examination. Users can also set up preconfigured views to track overall cost trends while customizing views to align with user requirements. AWS Cost Explorer has some incredible features that assist you in comprehending and managing your spending. Data Exploration: Access AWS Cost Explorer directly from the Billing and Cost Management console. With the ability to filter and group data by dimensions such as service, usage type, region, and tags, you can dive deep into specific areas and identify key cost drivers. Report Generation: Build custom reports to monitor your costs and usage trends over time. These reports can be saved and scheduled for regular updates, offering continuous insights into your spending patterns. Costs Forecasting: Leverage Cost Explorer's forecasting feature to predict future expenses based on historical data. The predictive faculty supports better budget planning and more efficient resource allocation. Businesses that do not optimize costs can face higher costs, inefficient use wastage of resources, and severe impact on profit. Therefore, AWS cost optimization is essential for utilizing cloud resources efficiently and cost-effectively, potentially saving money for businesses in the long run. The benefits of cost optimization include lowering costs, offering predictability, preventing unforeseen expenses and cost overruns, and enhancing resource utilization, which drives business growth and sustainability.

Looking to up your game in Business Intelligence and Data Visualization? Want to explore the endless possibilities of Tableau, showcase your work, and get valuable feedback from like-minded peers? Then TUG Meetup is the right event.

The Second TUG Cairo Meetup

Beinex has sponsored the second Tableau User Group Meetup in Cairo, and we are glad to initiate a thriving community that promotes learning, growth, and networking. TUG provides the perfect platform to connect with others, exchange ideas, and gain inspiration whether you're a beginner or an expert in Tableau. At the second TUG Meetup Cairo, the participants got the opportunity to showcase their work, receive constructive feedback, and be part of a community that shares their passion for data visualisation.

Date: March 11, 2023
Venue: MQR, The GrEEK Campus Downtown, Cairo Governorate 11513
Time: 2 PM to 5 PM GST.
We had a great discussion with the following eminent speakers on the topics enlisted below:

#	Speakers	Designation	Topics
1	Khaled Hoza	Data Analyst Team Lead, Fawry	Analyse the performance of your customer retention strategy
2	Martina Ghali	Senior CRM Specialist, Cartona	Overview of digital marketing and get it visualised in Tableau
3	Ahmed Ismail	NLP Engineer, Agolo	Visualise your text analytics in numerous ways

Abdelaziz Mahjoub, Data Analytics Lead Consultant, Beinex was the leader of the TUG Cairo Meetup. He is the first and only Tableau Public Ambassador in Egypt and MENA Region, Tableau public featured author, 1x #VOTD, and Tableau certified associate with practical and academic knowledge of Essential Design Principles techniques.

Initiating a New Tribe that Helps to Grow

[sc name="quote" quote="“Since my early days in the field, I struggled to get help on how to start, a lot. I know how hard it’s to fully understand a certain technology or a tool to a mastery level without proper guidance and mentoring. Here in Cairo TUG, I try my best to help other people not to find themselves in my position back then. That’s why I started Cairo TUG, to build a strong community that everyone can rely on to get help and find the proper guidance”" author="Abdelaziz Mahjoub, Lead Consultant, Analytics at Beinex Consulting,
He is the master brain behind the Cairo TUG."][/sc]