Supply Chain Expertise and GenAI Insights

I have worked for Microsoft Xbox supply chain and UPS. Now If I have to add a GenAI lens to my past. Here are my perspectives.

At Microsoft's Xbox division, I managed both forward and reverse supply chain logistics, covering repairs, refurbishment, and warranty services. By integrating GenAI, we can significantly enhance product defect detection, implement targeted product improvements, and streamline supplier communications, elevating overall efficiency and product quality.

At UPS, I collaborated with the team to develop and pitch AI/ML strategies for third-party logistics, focusing on operational efficiencies, visibility, planning, forecasting, and optimization. With the infusion of GenAI, the potential applications could extend to sophisticated chatbots and digital assistants, further refining company policies, fostering broader AI adoption, and enhancing supplier communications.

I am committed to pushing the boundaries of supply chain management with innovative GenAI-driven solutions and look forward to collaborating with startups/product companies in this space. If you are an SMB, have historical data, looking to onboard in AI, Let's connect sivaram[at]phygitalytics.com. 

Keep Learning!!!

    

Insights, Innovations, and Lessons: Exploring Computer Vision and Generative AI Hybrid solutions

Hoping to share more insights on below Vision + GenAI Journey

1. Captivating Success: Harnessing Vision and Generative AI to Mitigate Food Waste

Unveiling a remarkable deployment where vision technologies, coupled with Generative AI, are being leveraged in a significant initiative to reduce food waste. A partnership between a renowned condiment brand and a leading technology provider exemplifies the powerful application of AI in environmental sustainability.

2. Dual-edged Experiences: Enhancing Product Details and Vision Technology Setbacks

This segment will delve into the mixed outcomes from integrating Generative AI in enriching product detail pages, and the limitations encountered with computer vision technologies. We’ll share an analysis of the decision-making processes in either developing custom vision models alongside Generative AI or opting for off-the-shelf solutions, outlining the key challenges and learnings from both paths.

3. Learning from Setbacks: Challenges in Vision for Skin Care Innovations

Not all ventures yield success, and in the explorative landscape of AI, the application of vision technology for skin care solutions has faced its own set of challenges. This case will reveal the hurdles faced during implementation and the pivotal lessons learned, emphasizing the importance of iterative testing and adaptive strategies in technology application.

Summary:

In wrapping up, the session will highlight the critical takeaways from the successes and setbacks observed in integrating computer vision and Generative AI across different industries. Attendees will gain a nuanced understanding of the practical applications, scalability issues, and strategic decisions crucial for leveraging these cutting-edge technologies effectively.

Keep Exploring!!!

Career Perspectives

Few key things that apply to me :)

• Build key relationships by networking and offering something valuable - ideas/playbook / POC.
• Fifth, select challenging projects that promise significant early wins, even if they require extra effort initially.
• Finally, at the end of the month, create a reflective document detailing your observations to better understand and improve your approach.

Ref - CAREER ADVICE: First 30 days as an exec.

Keep Exploring!!!

Windows Ad - 1988

 

Keep Exploring!!!

Evaluating Common Sense AI Frameworks vs. Business-Driven Realities

The following compares two contrasting approaches: a common sense AI framework that focuses on ethical ideals, and the business-focused reality that often prioritizes profitability and market demands:

People-Centric vs Business-Centric Approach

• Common Sense AI: Prioritizes individuals' needs and well-being.
• Business Reality: Focuses on enhancing business profitability and operations.

Encouraging Learning vs Driving Engagement

• Common Sense AI: Advocates for learning and the personal development of users.
• Business Reality: Strives to increase user engagement which often translates to increased revenue.

Facilitating Connections vs Forming Opinionated Groups

• Common Sense AI: Aims to help people create meaningful connections without bias.
• Business Reality: May encourage the formation of groups based on strong opinions, which can boost platform activity.

Trustworthiness vs Promotion of Varied Information

• Common Sense AI: Committed to being trustworthy in the information it provides or promotes.
• Business Reality: May distribute all types of information, irrespective of accuracy, to cater to diverse user demands.

Privacy Defense vs Utilizing Data

• Common Sense AI: Upholds the privacy of users as a fundamental principle.
• Business Reality: Sometimes utilizes user data without explicit consent to maximize business opportunities.

Safety for Minors vs Conditional Apologies

• Common Sense AI: Ensures the safety of children and teens as a priority.
• Business Reality: Focuses on rectifying issues only when necessary to maintain public image while keeping business priorities intact.

Transparency and Accountability vs Limited Oversight

• Common Sense AI: Maintains high levels of transparency and holds itself accountable to stakeholders.
• Business Reality: Oftentimes finds methods to collaborate or operate without stringent checks, prioritizing flexibility and operational efficiency.

Keep Exploring!!!

AI in Beauty / Skin Care

Congratz Khusbhu, Mastering Vision + Beauty domain takes well calculated approach. This implementation is great example.

Selecting the right solution approach is the key

• Decision of Build vs Buy Model 
• Market testing
• Model Evaluation
• Data Compliance

Build is an expensive route in this case as it needs Deep Expertise in Vision plus Data Collection to culture.

Build vs Buy Solutions

• haut 
• Perfect Corp 
• Skintelligent  
• ModiFace 

Keep Exploring!!!

AI Solution Strategy Perspectives

Over the last few days, there has been intense discussion around product architecture, design, and adoption. Here are some key points:

• Balancing the adoption of Large Language Models (LLMs) with considerations of cost, consistency, and latency.
• The architecture should be flexible enough to allow plug-and-play integration of different models.
• Every solution must have some differentiation, value add, or a "secret sauce".
• AI tends to perform best when combined with human input (AI + Human in the loop).
• From time to time, use a "convince with code" approach to demonstrate solutions.

Keep Exploring!!!

OpenAI - Prompt King - Prompt Usage Patterns

The interesting thing about OpenAI is the prompt history. The information below is a gold mine:

• Commonly used prompts and their responses.
• Ranking responses based on user feedback.
• Distribution of prompts across different domains. (Health, History, News, Tech)
• Caching of prompts and answers for quicker access.
• Low latency approach to handling cache versus read operations.
• Asynchronous processes involved in domain detection, intent detection, and retrieval.
• Various combinations of indexes are used to optimize searches using golden data, cached data, summary data, and raw data.

Keep Exploring!!!

Good Paper Read - THE LANDSCAPE OF EMERGING AI AGENT ARCHITECTURES FOR REASONING, PLANNING, AND TOOL CALLING: A SURVEY

THE LANDSCAPE OF EMERGING AI AGENT ARCHITECTURES FOR REASONING, PLANNING, AND TOOL CALLING: A SURVEY

Key Summary notes

• AI agent architectures are either comprised of a single agent or multiple agents working together to solve a problem.
• Agent Persona. An agent persona describes the role or personality that the agent should take on, including any other instructions specific to that agent
• ReAct. In the ReAct (Reason + Act) method, an agent first writes a thought about the given task. It then performs an action based on that thought, and the output is observed
• Reflexion. Reflexion is a single-agent pattern that uses self-reflection through linguistic feedback

Dify is an open-source LLM app development platform

Beyond LLMs: Agents, Emergent Abilities

1. Agent is able to split / create smaller subtasks

2. Persona can be set for the agent, Few shot examples supplied to get the context

3. Multiple agents created for different purposes

4. Custom Agents / Agent to Agent communication etc..

Keep Exploring!!!

Career Blues and perspectives

My selected list from below bookmarked articles

• Try to become a new person every 3-6 years.
• Try to solve big problems as fast as possible. 
• Adding a little bit of extra productivity to every day is great advice. The challenge can be finding the time, which means you need to subtract time from some other activities.
• I write only when inspiration strikes. Fortunately it strikes every morning at nine o'clock sharp — W. Somerset Maugham
• Focus more on my own future rather than on what others think.
• Set clear short-term and long-term career goals.
• Enjoying programming and having the time outside of work. When it's a passion it all becomes a lot easier.
• Effort and consistency trumps all
• I personally believe anybody can find success if they just focus on the journey and the joy of coding, make sure they show up and participate, be present, stay curious and playful, and don't expect any rewards for their efforts. The real reward is having fun and feeling fulfilled while you're creating something.

Bookmarks

• Measuring personal growth
• The unreasonable effectiveness of just showing up everyday
• Outstanding Programmers
• My 25-Year Engineering Career Retrospective

Keep Exploring!!!

Different stages of ML / DL Learning

I want to learn ML -> Take a course 
I know the basics from the course -> Try the code examples 
I tried but I don't know what's next -> Find a use case 
I found a use case -> Collect data 
I collected the data -> Model the ML problem 
I built an ML model -> Create an API to consume it 
I built an API -> Dockerize it 
Is the API scalable? -> Check options such as serverless functions, Endpoint providers like Anyscale / SageMaker Endpoints, GCP, Azure Inferencing 
I deployed the model -> Version your models using MLFlow 
When to update -> Audit / Track data 
What tools to learn -> Align with what your organization uses and cloud vendors

Learn to walk before you try to fly. Everything is incremental learning. Keep going!!!

Keep Exploring!!!

Good vs Bad Leaders

From my 2 decade observations,

• Title <> Experience 
• Title <> Knowledge 
• Good failures can be Good knowledge experiences but people without Titles

I have worked with all ages 20's, and 30's. My peers.

• With my juniors, I give them context, references, and How I would do in their situations.
• With my seniors, I get only what to be done, I don't mind, work comes to me what they can't do
• With my peers, When asked for input, I share the same
• You meet people with Ego without knowledge. Not knowing and pretending is more dangerous.

Be an Empathetic Leader and Manager. Be accountable for your work

Keep Exploring!!!

Good Read - Is mid-career unemployability the big issue no one's talking about?

Good Read from thread - Is mid-career unemployability the big issue no one's talking about?

Some key points from the thread

• Hit refresh button and run your own startup in any sector build new skills explore new opps
• Help others as much as possible.
• I agree with you and I believe most experienced professionals are actually wasting their expertise by working as an employee and paying lakhs in income taxes
• I think everyone who is good at what they do and are over 40 should learn to become a consultant.
• There is a lot of value in being able to bring people together, solving problems holistically and from first principles
• Building your career progression to have higher impact, bigger outcomes, critical decision making and mentoring the next tier is the way to go. 
• The skillsets you have acquired through your career is the USP that you have
• Growth in terms of skills, decision making, managing complexity or mentoring people
• Putting a self-paced learning system in place, ranging from making time, choosing learning goals, and modes that work for you
• Embracing lifelong learning opportunities, staying updated with industry trends, and acquiring new skills can help seasoned professionals remain relevant and valuable in a competitive landscape. 
• Most importantly, one must maintain a strong professional network and stay visible in their industry to stay top of mind among one's peers and discover new opportunities.

Keep Exploring!!!

Data Science & Data

Every project is a learning experience. Data science is based on "Data". Working with no data, less data, or encrypted domain knowledge with minimal data has been challenge over the past 4 years. Yet, even when data is plentiful, there remains a balancing act between leveraging it effectively and mitigating trust issues, as collaboration can sometimes be overshadowed by the scramble for credit. Everyone wants to work on a model, not on data, the old google paper still comes into their eyes :). The current trend is to train large language models (LLMs) on uniform datasets, yet this approach glosses over an important truth: no dataset can capture the full spectrum of reality. Issues such as digital poverty, underrepresentation, and inherent biases are embedded within the data we collect. Without addressing these challenges, solutions can be superficial and short-lived. Moving fast with a lot of guardrails is essentially a band-aid, not a solution. Take a step back and balance data vs model. Build something that lasts forever not for paychecks!!!

Keep Thinking!!!

Why we don't see good AI / ML work

Why we don't see good AI / ML work. Good AI / ML work depends

• Choosing a use case with a strategic and AI-focused approach. Selecting the use case that has a balance of vision/strategy / applying AI lens
• Ensuring access to adequate data for training and deployment
• Garnering robust business support
• Acquiring or developing tools to deliver meaningful AI/ML contributions

Struggling with your AI strategy? Let's connect and navigate it together.

Keep Exploring!!!

Anyscale Endpoints discussion

Step #1 - Anyscale signup

Step #2 - Notebook for Deploying Diffusion models

Step #3 - Deploying Service Command

Step #4 - Service Deployment

Code Example - 

Keep Exploring!!!

2024 - Year of Opportunities / Lessons / Learning's

My memorable moments/projects/achievements.  

• Build vision capability
• 2 Granted Patents
• Won Vision solutions for Retail, and FMCG customers
• Built products in a consulting role (Some failed / some worked)
• Rewrote warranty for 220million consoles in Microsoft

Next Steps

• Teaching + Deep Dive + Part-time is my goal for sustainable health and learning
• Pick and select a few things and deep dive and build a point of view

Keep Exploring!!!

Data is not the new oil - Alexandr Wang

Great Talk, Lot of good insights

• Oil is a commodity and everyone has same access to oil
• Not all data has same level of value
• Every type of data fintech / insurance / healthcare has different 
• Data has multititude 
• Multiple frameworks / thoughful strategy to stitch data
• Building block for next move is quality code / automated code
• Earliest is Autonomous cars
• Raw data to labelled data is First step towards quality data
• Data = New Oil, Scale = Refinery
• Most capabilities are taught by large scale data
• Data Engine = Refinery for data
• Teach a model how to access one answer better than other with a bunch of examples

Opportunity for Enterprises

• Total data available - 99% - Private
• Messages / Emails will never end up on internet
• Enterprises have a lot of unused data
• Tune General purpose model for Enterprises
• Customer care / Legal Apps
• Focus on big problems that matter to your business

Questions

• Unique Data Assets
• Better than Anyone else 
• Unique capabilities / Differentiators
• Cost Reduction / Customer Care / Optimization
• AI = Productivity enhancer
• Meaningful chunks of work
• Health care / Financial Services
• Data & Compute Limiting Factors
• AI <> Replacement for humans
• Economically viable human systems
• There is no future here 2 years back vs This is a threat
• Model improvement is going to get better
• Need broader cooperation
• Inequality with jobs / upskilling with new jobs
• AI misuse is punished / handled severly
• Testing and Evaluation for Systems and use case
• Fit for purpose vs primetime
• FDA for drugs similar regulation options, Apps approved after scruitiny
• Public evaluation of models
• Testers in public / private / regulators
• Ideas + Accountability decentralized

Ref - Alexandr Wang: 26-Year-Old Billionaire Powering the AI Industry

The Truth About Building AI Startups Today

• GPT Wrappers
• AI Agents 
• High Beta Opportunities
• Idea Maze
• Once in a Life time opportunity
• AGI / Multimodal / Videos
• Workflow Automation
• RPA - Search/ Form Filling
• Sweet Spot - Pivot LLMs Automate Government Contract
• Dev tool companies
• Fine Tuning LLM Models
• Something more than Finetuning
• Customize to private datasets (Healthcare / FinTech)
• Cybersecurity to Cloud -> Cybersecurity for LLM
• LLM to data access Mapping
• Purpose trained models / Run Locally models
• Prototype with close source LLM Models
• Collect data in parallel for domain context
• Partner and build private models
• Closed source AGI = Monopoly / Dangerous
• AI Ethics + Regulation + Measuring it

My Take

• Prototype with close source LLM Models
• Collect data in parallel for domain context
• Partner and build private models

Product #1 - Syncly

• With AI feedback analysis, Syncly instantly categorizes feedback and reveals hidden negative signals. 
• Centralize all your feedback and take proactive actions based on real time insights to elevate your five-star customer experience.

Product #2 - Cradle

• Protein engineering without the guesswork

Keep Exploring!!!

Transformer Walkthrough

Session #1

Session #2

Keep Exploring!!!