Anwar Zahid Khan

What can Enterprises do to have AI available to everyone in the org and govern/manage it centrally? 🚨 Pick a platform that can have all your Orgs Automation, AI models at one place. 🚨A platform that offers complete governance including ModelOPs, Performance, Drifts and Visibility. IBM GenAI vision incorporates the above two themes in WatsonX foundation. It’s a simple architecture to innovate and manage GenAI projects. DM for more… #watsonX #watsonXorchestrate #aiautomation #aigovernance

6

Deploying updated models requires careful planning to avoid disruptions to users and ensure a smooth transition. Several strategies can be employed to test and deploy the new model with minimal risk. A/B Testing allows teams to compare the performance of the old and new models by randomly assigning users to each model. Canary Releases involve deploying the updated model to a small subset of users before rolling it out to the entire user base. In a Shadow Deployment, the new model is deployed alongside the existing model, but its predictions are not shown to users. Get more of these useful insights in my book "Machine Learning for Product Managers" - https://a.co/d/0gSzUnl Get your copy today. Available exclusively on Amazon. #AIForProductManagers #ProductManagers #ArtificialIntellegence #ProductManagement #PM

7

2 Comments

Why Enterprises need Productized solutions for realizing AI Use Cases? 🚨 I am an Enteprise Architect working with my team to build an Agentic architecture which will look something like below. 🚨As you can see there are a number of elements in this architecture that are loosely coupled to get the end response. 🚨 Now think about maintaining the Day 2 operations of these individual elements as you scale your use cases and expand across several geographical regions. 🚨 Can you think about the nightmare of managing a Sev1 in this architecture on TOP of your non AI apps that are running in Prod? 🚨 IBM WatsonX Orchestrate provides a Productized version of AI agent creation on top of creating non Agentic Assistants by combining AI and Automation. The product encapsulates the operations teams by managing a 100 components and gives “one” place to look for issues. AI architectures evolve by the hour, Big Enterprises must think about how much “custom” they need to go. The more custom it is, the more SRE nightmare it is for Day2. Day2 easily goes under the covers as it’s always an after thought, lead every design with Day 2 in mind for better decisions. Check the video below on WXO handing AI agents as a product. #watsonx #watsonxorchestrate #aiagents #ibm https://lnkd.in/guxzCWsN

8

Congrats on the insightful article on Gen AI and GraphRAG by Maruti C and Michael Hunger ! ! Their thorough explanation offers valuable perspectives on this adanavnce technology. Great job on publsihing this , Read the article here:

4

1 Comment

In today's data-driven world, understanding data flow is essential for maintaining quality and compliance. Data lineage, tracking data from origin to destination, is crucial. My latest blog explores how Marquez enables detailed data lineage in Airflow and Spark ecosystems. Discover how we fetch data from SQL Server to a MinIO data lake using Iceberg tables, then to a PostgreSQL data warehouse and data mart. See how Marquez visualizes lineage at table and column levels. #DataManagement #DataLineage #Marquez #DataEcosystems

26

1 Comment

The Learning Machine and an AI scientist copying IT This is a concept highly spoken by Charlie Munger and references in Gautam Baid, CFA book The Joys of compounding. I sincerely agree with the concept, valuable reading just makes you smart and the human brain is a wonderful creation of God in human beings. The smarthome prototype that I did to help my kid understand the mechatronics, IOT in just two days was a eye opener for me on how fast our brain can think and work with clarity if you control unmindful thoughts and control your brain. It seems the AI scientist listened to the idea of Uncle Charlie’s The Learning Machine and applied the concept to AI by a concept known as unsupervised Learning to make Mr.AI smart as a human being. 🤓

3

1 Comment

Had an amazing time at last week's hackathon at Zensurance investigating approaches to simulating an AI agent. (Simulating is a key word here as there is no true reasoning with the tech at hand). Approach - 🔝 At top level, a guidance prompt to direct the agent one way or the other 🖇 A knowledge graph to store knowledge of relationships between objects 🚀 An LLM to take human input and convert it into knowledge graph or other queries 🎁 A wrapper program with a list of functions to call Observations - ✔ This approach will work in situations where the list of actions the agent can pursue is known beforehand ❌ The KG (https://neo4j.com/) was useful to store relationship knowledge but the task of loading it with data was cumbersome - the KG way of querying was not very intuitive and takes getting used to 💠 Learnt a little too late that Microsoft had released a RagGraph package that helps load a graph database from text - will investigate this the next time ✔ OpenAI performed well in translating human questions to graph queries and vice versa - so far it has been the best performing of the options investigated across other investigations (https://lnkd.in/gvfB5RqG) ⌛ There are distinct latency issues if both the LLM and the KG are on separate cloud instances - need to investigate whether a local KG can be built so one of the jumps can be eliminated 🗒 The approach is extensible in terms of adding functions - it should be possible to make the expansion of the functions list configurable, such that new functions can be deployed without having to re-deploy the entire code - this needs investigation as well 🕸 Another important thing learnt - how to quickly deploy a web application - having investigated a handful of options, found python based Flask (https://lnkd.in/gtMzQqsZ - see https://lnkd.in/gNJ5_Xkm for amazing tutorials) to be the most powerful yet intuitive and easy to learn - a great way for data professionals to build web apps (checkout their template feature). [Once you've understood how to deploy using flask, not too much of a leap to jump to the more powerful django, either!]

10

Interesting thoughts from Junta N. about #AI use cases in #insurance.

19

Lately I've been learning more about #SQLMesh from Tobiko Data. It is a data transformation tool similar to #dbt, but with an architecture designed to optimize your cloud spending. While it may be new, SQLMesh already has a lot of robust features that make it a strong contender. You can read more about my thoughts on it here. https://lnkd.in/euHefGsz

27

ChatGPT told me that I am going to become a data scientist. Training specialized large language models for applications like cybersecurity and insurance tech involves many core data science skills, including data collection and preparation, model training, statistical analysis, and perhaps most importantly, the interpretation and application of model outputs. Here's a closer look at the key aspects: 1. Data Expertise: Understanding and manipulating large datasets is a fundamental part of data science. This includes cleaning, preprocessing, and ensuring the quality and relevance of the data used for training models. 2. Modeling Knowledge: Knowing how to choose, train, and fine-tune models—whether they are pre-trained models like those from Hugging Face or custom models you've built from scratch—is crucial. This involves understanding machine learning algorithms, neural networks, and possibly deep learning frameworks like TensorFlow or PyTorch. 3. Statistical Analysis: Assessing the performance of your models through statistical measures and being able to interpret these results to make informed decisions is essential. 4. Domain Expertise: For each application (cybersecurity and insurance tech), having or developing domain-specific knowledge that informs the data and modeling choices is vital. 5. Ethical and Legal Considerations: Being aware of and adhering to ethical and legal standards, especially when handling sensitive data or deploying models that make significant decisions, is part of the data science discipline. You're actively engaging in these activities, you are indeed performing the role of a data scientist. Data science is a broad field with applications across virtually every industry, and it encompasses a range of activities from data management to advanced analytics. Training and deploying models as you described puts you solidly in the realm of data science work. Whether you're considered a "data scientist" by title may depend on the context or your professional environment, but certainly, the skills and tasks you're taking on are at the core of what data scientists do.

6

🚀 Building Scalable, Maintainable, and Resilient Applications? Start with the 12-Factor App Methodology! 🌐 In today’s cloud-native world, creating applications that are scalable, maintainable, and ready for the challenges of dynamic environments is more important than ever. Whether you're working with microservices, cloud platforms, or modern DevOps practices, the 12-Factor App methodology offers a proven framework for success. I've just published a comprehensive blog post that dives deep into each of the 12 principles, explaining how they can transform your approach to application development. From codebase management to treating logs as event streams, these best practices are essential for anyone looking to build robust, cloud-native applications. 🔍 Curious to learn more? Check out my latest blog where I explore each factor in detail, provide book recommendations, and share insights on how to apply these principles in your projects. Read the full blog here: https://lnkd.in/efA6vUgP #CloudNative #DevOps #SoftwareDevelopment #Microservices #12FactorApp #SaaS #CloudComputing #Scalability #Resilience #BestPractices

18

4 Comments

Interesting Gen AI perspective on AI and Culinary What impact does South Indian cuisine have on culinary exploration? South Indian cuisine significantly contributes to culinary exploration by offering a rich tapestry of flavors and techniques. Nathan Nagarajan's experiment with Gen AI and South Indian flavors, like the sambar dish, highlights the potential for AI to innovate within traditional cuisines, creating new recipes that could match or even surpass those of seasoned chefs. This fusion of technology and traditional cooking opens up exciting possibilities for culinary creativity and innovation. How does AI influence culinary creativity? AI influences culinary creativity by enabling the exploration of novel flavor combinations and recipe creation, as demonstrated by Nathan Nagarajan's experiment with South Indian cuisine. It offers tools for chefs to innovate beyond traditional methods, potentially leading to groundbreaking culinary developments.

2

𝑭𝒂𝒊𝒓 𝑺𝒐𝒖𝒓𝒄𝒆 𝒗𝒔. 𝑶𝒑𝒆𝒏 𝑺𝒐𝒖𝒓𝒄𝒆: 𝑾𝒉𝒂𝒕 𝑬𝒏𝒕𝒆𝒓𝒑𝒓𝒊𝒔𝒆𝒔 𝑵𝒆𝒆𝒅 𝒕𝒐 𝑲𝒏𝒐𝒘 As enterprises increasingly rely on open-source software to drive innovation, it's important to understand the nuances of alternative licensing models like Fair Source. While open source is built on the principle of free use, modification, and distribution for anyone, Fair Source takes a slightly different approach: 𝐎𝐩𝐞𝐧 𝐒𝐨𝐮𝐫𝐜𝐞: Completely free for everyone to use, modify, and distribute, governed by licenses like MIT, Apache, or GPL. 𝐅𝐚𝐢𝐫 𝐒𝐨𝐮𝐫𝐜𝐞: Free for a certain number of users (typically 10-25), but once that limit is crossed, companies must pay for licenses. The source code is still available, ensuring transparency, but the model encourages compensation after reaching certain usage thresholds. 𝐈𝐦𝐩𝐚𝐜𝐭 𝐨𝐧 𝐄𝐧𝐭𝐞𝐫𝐩𝐫𝐢𝐬𝐞𝐬: For large organisations, Fair Source provides a middle ground—allowing teams to try, test, and innovate with software before committing to a paid model. It's a great option for enterprises that want to leverage the benefits of open code with the understanding that growth will lead to a paid relationship. 𝐀𝐝𝐯𝐚𝐧𝐭𝐚𝐠𝐞𝐬 𝐟𝐨𝐫 𝐄𝐧𝐭𝐞𝐫𝐩𝐫𝐢𝐬𝐞𝐬: 1. 𝐂𝐨𝐬𝐭 𝐂𝐨𝐧𝐭𝐫𝐨𝐥: Start small and scale, with costs only incurred after a threshold is crossed. 2. 𝐒𝐮𝐩𝐩𝐨𝐫𝐭 𝐚𝐧𝐝 𝐒𝐮𝐬𝐭𝐚𝐢𝐧𝐚𝐛𝐢𝐥𝐢𝐭𝐲: Supporting creators while still accessing cutting-edge technology. 3. 𝐓𝐫𝐚𝐧𝐬𝐩𝐚𝐫𝐞𝐧𝐜𝐲 : Enterprises still get access to the code, ensuring they can review and tailor it to their needs. As more companies embrace hybrid models like Fair Source, it may become a cornerstone for sustainable innovation in enterprise tech. https://fair.io/about/ #opensource #fairsource #enterprisetech #softwarelicensing #innovation

18

2 Comments

Thrilled to be part of Intel’s mlcon3! 🚀 Honored to give a presentation at #mlcon3 on the latest advancements in #ArtificialIntelligence and #MachineLearning. It is an honor to be among the thought leaders and innovators at this hybrid event. 🤖 Diving into the transformative power of AI & ML, I explored how these technologies are reshaping our industry and enhancing human capabilities. 🌐 A big shout-out to the organizers for creating a platform where knowledge meets innovation! 🔗 Stay tuned for more updates and feel free to connect if you’re interested in collaborating or learning more about AI and ML. #AI #ML #Innovation #Technology #mlcon3 Allstate

51

1 Comment

For more than a decade of my professional career, I witnessed how Data & AI technology can create advancements and value for business, specifically in financial services. For leaders and decision takers its important to understanding how investments in AI technology benefit business outcomes! ❔Are you seeing ROI from your company's #AI in #Finance? Today KPMG International published an excellent Global research report providing valuable market insights on #AI-empowered #finance function. Our latest global AI study was conducted among CFOs and finance executives across 2,900 companies spread across 23 countries and territories and 6 industries and found that the companies surveyed are rolling out AI across wider areas of finance, including accounting, financial planning, treasury management, risk management, and tax management. The key findings from the study include: ❇️71% of organizations are using AI within finance operations and 41% using AI to a moderate or large degree, as AI adoption is increasing globally. ❇️99% of companies have plans to use Gen AI for financial reporting, compared to 94% in our previous survey. ❇️57% of leaders say ROI is exceeding their expectations, compared to 29% of implementers and beginners. ❇️Finance executives want more AI communication from their auditors. Only 15% of leaders say their auditor communicates frequently with them about AI — but 59% would like them to Financial reporting is the most common usage area for AI, but this is expanding to include accounting, financial planning, treasury management, risk management and tax management. ➡️Find the report and related insights https://lnkd.in/dNJ2R4fk KPMG Switzerland KPMG UK David Rowlands Leanne Allen Florian M. Häller Christoph Steiner Thierry Kellerhals Rene Vader Martin Thomas Benedikt Lindheim Fabio Sabatini

39

1 Comment

Guidewire's Umang Jain and Yoganand Ghati recently collaborated on a Medium blog post, "Organizing Work for Simplicity and Improved Collaboration," to discuss the on-premise-to-cloud journey that Guidewire is making and how program management in its cloud platform engineering team has been instrumental in enabling teams and stakeholders to deliver consistently and predictably. #GuidewireCloud #cloudcomputing #insurance #innovation #insuranceinnovation #insurtech #insurancetechnology #propertyandcasualtyinsurance #development #engineering #software

2