Building Resilient Supply Chains with AI - Webinar Notes

Key Notes

• Covid-19, Squez Canal 
• New markets, New products
• Demand for more WFH essentials
• Long term forecast and short term finetuning

• Demand Sensing - Adapting to fluctuations
• No Historical data
• Excel Models, Rules-based models
• History keeps changing every few months

• Macro economic factors
• Consumer price index
• Producer price index
• Unemployment claims
• Mobility Data
• Inflation

• Country - Dept - Store - Sku - Sales
• Department - Product - Delivery Time Delay
• Deploying all this models, Real time deployments
• SEIRD Models
• SEIRD Model Paper
• Epidemic analysis of COVID-19 in China by dynamical modeling

H2O.AI also seems to be one more kubeflow customization

Google Vertex also seems a combination of google AutoML + Kubeflow Customization

Keep Exploring!!!

Vision - Webinar Notes - Life of Vision Project - OpenCV - Labelmg - Object Detection - Tensorflow Lite - KfServing - MySQL

Some talks summarize all of your work, plus talk on those aspects which you did implicitly part of your work. Nice Talk to connect with Vision Work :)

Key Notes

• Workflow for Image Solutions
• Dataset preparation
• Data Annotation
• Training / Benchmark
• Pre-trained + Transfer Vs Custom Model
• Metrics for benchmarking

Data Collection

• Data set type
• Streaming or Image
• Data formats
• Single image / frames
• Video - Frame - Feed model
• Image Resolution / Frame rates sampling
• Reduce frame rate to support more streams
• Preprocessing work
• Crop noisy areas
• Select areas of interest
• Data Generation
• Data Augmentation
• Simple techniques including vision tricks - rotation, transformation, different angles
• GAN / Synthetic data generation techniques

Data Annotate 

• Annotate / Review
• Validate with SME
• Bounding boxes / Segmentation / Labels
• Single / Multiple objects / Classes
• Occlusion, Light, settings
• Partially available surfaces
• Fine-grained annotation or not
• Data set representation against bias
• Coverage of possible classes
• Models for Day time vs Night time

Model Training

• Segmentation / Custom Detection
• Post-processing
• Transfer Learning

Model Optimization

• Prune / Quantize
• Inference Engines
• CPU / GPU / FPGA

Benchmark

• Testing on Deployable hardware
• Number of endpoints
• Load vs Response
• Re-annotate / Re-train
• Ensemble or Single Model

Deploy

• Edge vs Cloud 
• Edge Server - Lite weight models
• Address based on the constraint, workloads for edge devices
• Hybrid approach both edge + cloud
• Model interface with application
• Storing Results in DB
• Real-time notification or just store

Model Monitoring

• Monitoring for data/accuracy of detections
• Pick low accuracy results / retrain them
• Capture when confidence is less than 50%
• Continuous re-learning 

End to End platform for this

Keep Connecting the Dots!!!

Anamoly Reads

Ref - Link

Summary

• Point Anomalies - Value is far outside the entirety of the data set
• Conditional Outliers - With respect to context, Same value may not be anamoly in another time 
• Collective Outliers - Set of 1 or more points that deviate from dataset

Telemanom 

Paper - Machine Learning for Time Series Anomaly Detection

Key Notes

• Clustering methods do not require the data to be labeled, making it a good fit for our unsupervised task. Very sensitive to outlier data points

Two-Step Process

• The number of clusters can be set to 2 (one anomalous and one normal)
• Summarized by taking averages across an interval of one hour
• Rolling Window Sequences

On the Nature and Types of Anomalies: A Review of Deviations in Data

Patent - FAST AUTOMATED DETECTION OF ( 56 ) SEASONAL PATTERNS IN TIME SERIES DATA WITHOUT PRIOR KNOWLEDGE OF SEASONAL PERIODICITY 

Key Notes

• Calculate Automatic correlation based on timeseries values
• Identify local maxima
• The seasonal trend identification module
• Data store for Normal data, Anamoly data
• Scoring module
• Human in loop feedback system

Sklearn Models for Supervised Anomaly Detection. Some popular scikit-learn models for supervised anomaly detection include:

• KNeighborsClassifier
• SVC (SVM classifier)
• DecisionTreeClassifier
• RandomForestClassifier
• Interquartile Range
• Isolation Forest
• Median Absolute Deviation
• K-Nearest Neighbours

More Reads

Detecting Spacecraft Anomalies Using LSTMs and Nonparametric Dynamic Thresholding

Machine Learning for Time Series Anomaly Detection

Anomaly Detection Learning Resources.

Anomaly Detection in Time Series Data Using LSTMs and Automatic Thresholding

SYSTEMAND METHOD FOR TRANSFORMING OBSERVED METRICS INTO DETECTED AND SCORED ANOMALIES

Keep Reading!!!

The Dark side of Analytics

Interesting read - When algorithms dictate your work: Life as a food delivery ‘partner’

This applies to all aggregators - OLA, UBER, etc..

Key Perspectives

The Trap

• The illusion of guaranteed income while the variable incentives seem attractive initially
• Incentives riddled with a bunch of terms and conditions

Chasing Dreams

• Competition to be top performers
• Physical and mental costs of driving to complete those targets
• Average of 10 hours of being on the road
• A job that demands a large amount of time

Eating outside has costed my health very badly. In a way, this is not a sustainable business model. Until 35 your body will not show any problems. The effects will come up after 35. 

This business model is a toll on workers and they are not prepared for any better next job. Doing a good job vs Live your day vs Work for a better tomorrow.

Sustainable business models vs Profitable business models vs Inclusive social growth is always a question.

I am scared if the algorithm starts measuring my time

• How long have you looked at PC
• How many lines of code written
• How many functionality worked

Life is too short to see everything through the same lens. 

Keep Thinking!!!

Transformer - Let's relearn

Transformer - Let's relearn

These topics come and on and off. I was able to catch up with sliding windows, CNN, RNN, LSTM. Then a bit of Transformers and also how does it work in vision too :)

AI / ML won't let us feel guilty you have to still learn the basics.

Paper - Attention Is All You Need

Key Lessons

• Representation of the sequence
• Intra-attention of sequence order
• Encoder-decoder structure
• Encoder - a sequence of continuous representations
• The decoder then generates an output sequence (Positional encoding)
• Multi-Head Attention consists of several attention layers

Unofficial Walkthrough of Vision Transformer

• Image is also pixels, learning pixel representations then the same encoding / decoding can be applied.

Transformers for Image Recognition at Scale

Key Notes

• Input image as a sequence of image patches, similar to the sequence of word embeddings
• The Vision Transformer treats an input image as a sequence of patches
• ViT can learn features hard-coded into CNNs (such as awareness of grid structure)
• Image classification with Vision Transformer
• Image classification with Vision Transformer

AN IMAGE IS WORTH 16X16 WORDS: TRANSFORMERS FOR IMAGE RECOGNITION AT SCALE

• Split an image into fixed-size patches
• Linearly embed each of them
• Add position embeddings
• Feed the resulting sequence of vectors to a standard Transformer encoder

Do Vision Transformers See Like Convolutional Neural
Networks?
pdf: https://t.co/5Yz5F2PZwO
abs: https://t.co/bpHO2rOYDv

find striking differences between the two architectures, such as ViT having more uniform representations across all layers pic.twitter.com/0KT0KE16f9

— AK (@ak92501) August 20, 2021

Do Vision Transformers See Like Convolutional Neural Networks?

• Lower half of ResNet layers are similar to around the
• lowest quarter of ViT layers
• Highest ViT layers dissimilar to lower and higher ResNet layers.

Keep Thinking!!!

Man vs Machines

Interesting Read - Link

Perspectives from Stylist

• We knew from the beginning we were teaching the algorithm
• In recent months, sending out boxes of clothing that were entirely selected by the algorithm
• It’s like we’re constantly making the algorithm better by fixing their mistakes

Phases of this Progress

• Stylist recommendation (Observed by ML)
• Model Training based on recommendation (Learnt by ML)
• The model recommends, stylist finetunes it (ML Leads vs Stylist Finetunes)

Other reasons are also cited - changes in flexible working hours, minimum working hours, etc..

I have the same perspectives about the Tesla humanoid robot, How many jobs it can automate/eliminate. Machines that will never get tired/frustrated. 

The interesting debate lives vs cost savings vs empathy. Hope we build machines balancing all these aspects. 

Keep Thinking!!!

Observations and Perspectives

Sometimes commonsense work better than data science

Best practices are #point-in-time best practices with respect to the #data, #technology, #time, availability of #skills, time to fix solutions. It is easier to create but difficult to #evolve from where we are to what we want to become.

Data Science #Skills = #DiversityThinking = #DatabaseSkills + #DomainPerspectives + #DataScience thinking hat + Possibilities with available #data #bigpicture

Keep Questioning!!!

Reads - Forecasting

Paper #1 - Hierarchical forecasting with a top-down alignment of independent level forecasts

Key Notes

• Deep learning forecasting approach N-BEATS for continuous-time series on top levels
• Tree-based algorithm LightGBM for the bottom level intermittent time series
• Split the time series into the non-zero component and stochastic component
• The lowest level of the hierarchy exhibits a strong intermittent pattern
• Upper hierarchy levels contain forecastable components such as the trend or seasonality aggregated by the lowest level

Paper #2 - Hierarchical Dynamic Modeling for Individualized Bayesian Forecasting

Key Notes

Models for personalized forecasting should be

•  able to incorporate predictor information such as price and promotions,
•  adaptable to time-varying trends, regression effects and unforeseen temporal changes,
•  interpretable and open to intervention by users and downstream decision makers,
•  fully probabilistic to properly characterize forecast uncertainties and allow formal model and forecast assessment under multiple metrics,
•  adapted to hierarchical settings, and amenable to automated, computationally efficient sequential learning and forecasting.

We define three household groups based on total items purchased over the course of the 112 weeks:

• Household Group 1: high spending and purchasing households
• Household Group 2: moderate spending and purchasing households
• Household Group 3: lower spending and purchasing households

Household Group, Proportion Return, Mean Spend, Median Spend, SD Spend

More Reads

Optimal Combination Forecasts on Retail Multi-Dimensional Sales Data

hts: An R Package for Forecasting Hierarchical or Grouped Time Series

Keep Thinking!!!

Forecasting Reads - Research papers - Retail

Paper #1 - An industry case of large-scale demand forecasting of hierarchical components

Key Notes

• Demand forecasting system of electronic components in manufacturing

1. Algos leveraged - 1) Adaboost, 2) ARIMAX, 3) ARIMA, 4) Bayesian Structural Time Series (BSTS), 5) Bayesian Structural Time Series with a Bayesian Classifier (BSTS Classifier), 6) Ensemble of Gradient Boosting (Ensemble), 7) Ridge regression (Ridge), 8) Kernel regression (Kernel), 9) Lasso, 10) Matrix Factorization from section VII (MF), 11) Neural Network (NN), 12) Poisson regression (Poisson), 13) Random Forest (RF), 14) Support Vector Regression (SVR).

• Techniques on 1) data pre-processing, (2) prediction, and (3) model selection 
• Symmetric Mean Absolute Percent Error (SMAPE) serves to evaluate the performance of the models

Paper #2 - Learnings from Kaggle’s Forecasting Competitions

Key Notes

• High-frequency series at weekly, daily, and sub-daily levels
• Frequency data in the form of weekly, daily and hourly data
• Three full seasonal periods were required at each frequency  i) complex vs. simple models, ii) crosslearning, iii) prediction uncertainty and iv) ensembling
• Walmart Store Sales and the Rossmann competitions
• Sales by store/department/week and store/day
• Forecasts of unit sales being required by product/store/day

Data Preprocessing

• Set NA or Negative values to zero.
• Remove time series with all zero values. 
• Remove leading zeros.
• To calculate the feature vectors, we use the R package feats
• Apply principal components for dimensionality reduction using the prcomp algorithm

• Most of the top performers used ensembles of global XGBoost models to create forecasts, but a few of them did include local XGBoost models as part of their ensemble
• Holidays and promotion, turned out to be essential for obtaining high performance in this competition
• Global ensemble models outperform local single models

Feature Extraction

• Day of Week
• Weekend
• IsHoliday
• Ispromotionday
• IsMonthEnd
• IsyearEnd
• IsQuarterEnd
• IsLocalHoliday
• WeekofYear
• Wolling Window
• Average of 2 - 3 - Weeks
• Moving Average Numbers
• Mean Every 2 Weeks
• Incremental Differences Everyday
• Adding Averages / Means - Weekly Average, Daily Average

Paper #3 - An Empirical Analysis of Feature Engineering for Predictive Modeling

Following sixteen selected engineered features:

• Counts
• Differences
• Distance Between Quadratic Roots
• Distance Formula
• Logarithms
• Max of Inputs
• Polynomials
• Power Ratio (such as BMI)
• Powers
• Ratio of a Product
• Rational Differences
• Rational Polynomials
• Ratios
• Root Distance
• Root of a Ratio (such as Standard Deviation)
• Square Roots
• Counts - count engineered feature counts the number of elements in the feature vector that satisfies a certain condition
• Statisticians have long used logarithms and power functions to transform the inputs to linear regression

Paper #4 - VEST: Automatic Feature Engineering for Forecasting

• sku,wkno,saleqty
• cluster and forecast
• DWT - Dynamic Time Wraping Metric for clustering timeseries

For offline Retail Stores my list of feature variables

Store level stats

• Date
• StoreId
• Items in Store
• Traffic Count
• Holiday / Festival
• Number of Item Categories
• Weather
• Out of Stock Items
• Cost of Products, The price value of SKU
• Promotional Offers / Seasonal Information
• Weather Information on Store on that Day
• Store operational timings
• Store Labour Details

Data Product Thinking

• With 20% more restock of this item, It might reduce 10% out of Stock, 5% improvement in Traffic (Instead of blind forecast provide a collated recommendation)
• With 20% reduction in tomorrow traffic, corresponding items or % of Sale Can be presented
• Multiple models will run behind these decisions to generate the recommendations

More Reads

• Feature Engineering for Mid-Price Prediction with Deep Learning
• A clustering approach to time series forecasting using neural networks: A comparative study on distance-based vs. feature-based clustering methods.
• The autofeat Python Library for Automated Feature Engineering and Selection
• AutoAI-TS: AutoAI for Time Series Forecasting

Keep Thinking!!!

Measure Experience from Projects / Domain / Versatility / Evolving Perspectives

Career perspective at different stages

• 10 years of Working on the Same project = 10 years of Experience?
• 10 years Experience in 1 domain vs 5 domains (Multi-domain exposure)
• Ability to Translate Bird's eye view to Prototype
• Map prototype to Implementation tasks
• Work on the storyline in mind than just near focus tasks
• Code with Clarity vs Code with limited visibility
• Code with Domain knowledge vs Code and refactor based on domain knowledge
• Partner with Business and Work vs Work and Rework again for business
• Familiarity of Technology vs Visibility of use cases vs Clarity of implementation vs Ability to explain in an implementation perspective
• Map trends vs Current architecture vs Time vs Focus 
• Always keep thinking from multiple diverse perspectives

Keep Thinking!!!

Interesting Reads - Books H1

Some of the books I was able to review in the last 6 months. We need to revise again and again and experiment.

• Bird's eye view
• 30 K Perspective
• 20K Perspective

I am still learning. I take time to build my perspective. Experience is a mix of learning, doing, knowing, connecting with industry experts. Always be open to learning - unlearn - relearn.

Books List for future reference

• O'Reilly - A Practical Introduction to Supply Chain
• O'Reilly - Agile Conversations
• O'Reilly - AI Blueprints
• O'Reilly - Architecture Patterns with Python
• O'Reilly - Beautiful Code
• O'Reilly - Bioinformatics Programming Using Python
• O'Reilly - Breaking Out: How to Build Influence in a World of Competing Ideas
• O'Reilly - Building Evolutionary Architectures
• O'Reilly - Change Your Life with CBT
• O'Reilly - Cloud Analytics with Microsoft Azure - Second Edition
• O'Reilly - Communicate to Influence: How to Inspire Your Audience to Action
• O'Reilly - Data Governance: The Definitive Guide
• O'Reilly - Data Lake Analytics on Microsoft Azure: A Practitioner's Guide to Big Data Engineering
• O'Reilly - Design Thinking for Training and Development
• O'Reilly - Designing Data-Intensive Applications
• O'Reilly - Digital Supply Networks: Transform Your Supply Chain and Gain Competitive Advantage with Disruptive Technology and Reimagined Processes
• O'Reilly - Empathy (HBR Emotional Intelligence Series)
• O'Reilly - Exam Ref AZ-303 Microsoft Azure Architect Technologies
• O'Reilly - Fluent Python, 2nd Edition
• O'Reilly - Focus (HBR Emotional Intelligence Series)
• O'Reilly - Fundamentals of Supply Chain Theory, 2nd Edition
• O'Reilly - Graph Algorithms
• O'Reilly - Hands-On Vision and Behavior for Self-Driving Cars
• O'Reilly - How Stitch Fix uses human-in-the-loop machine learning for personalization
• O'Reilly - How to Persuade and Influence People: Powerful techniques to get your own way more often
• O'Reilly - Influence and Persuasion (HBR Emotional Intelligence Series)
• O'Reilly - Influence in Action: How to Build Your Conversational Capacity, Do Meaningful Work, and Make a Powerful Difference
• O'Reilly - Kubernetes in Action
• O'Reilly - Learning Python, 4th Edition
• O'Reilly - Linear Programming and Resource Allocation Modeling
• O'Reilly - Logistics Management
• O'Reilly - Machine Learning Design Patterns
• O'Reilly - Metaheuristics for Logistics
• O'Reilly - Nature-Inspired Optimization Algorithms
• O'Reilly - Practical Git: Confident Git Through Practice
• O'Reilly - Practical Machine Learning for Computer Vision
• O'Reilly - Practical MLOps
• O'Reilly - Practical Statistics for Data Scientists, 2nd Edition
• O'Reilly - Purpose, Meaning, and Passion (HBR Emotional Intelligence Series)
• O'Reilly - Python: Master the Art of Design Patterns
• O'Reilly - Resilience (HBR Emotional Intelligence Series)
• O'Reilly - Self-Awareness (HBR Emotional Intelligence Series)
• O'Reilly - Success in Programming: How to Gain Recognition, Power, and Influence through Personal Branding
• O'Reilly - Supply Chain and Logistics Management Made Easy: Methods and Applications for Planning, Operations, Integration, Control and Improvement, and Network Design
• O'Reilly - Supply Chain Management and its Applications in Computer Science
• O'Reilly - Supply Chain Management For Dummies
• O'Reilly - Supply Chain Optimization through Segmentation and Analytics
• O'Reilly - The Azure Cloud Native Architecture Mapbook
• O'Reilly - The Cloud-Based Demand-Driven Supply Chain
• O'Reilly - The Science of Influence: How to Get Anyone to Say "Yes" in 8 Minutes or Less!, Second Edition
• O'Reilly - Visual CBT: Using pictures to help you apply Cognitive Behaviour Therapy to change your life

Bookmarks for future reference!!!

Keep Thinking!!!

Research Paper Reads - Logs Monitoring

Ideas need a birds-eye view of the landscape to understand existing work. Papers are the only way to understand that. Bookmarking few notes for future reference

Paper #1 - Log-based software monitoring: a systematic mapping study

Key Notes

• The Lifecycle of Log

• Possible components would be Elasticsearch, Logstash, and Kibana
• Kibana provides an interface for visualization, query, and exploration of log data

• LOGGING - 1) empirical studies on logging practices, (2) requirements for application logs, and (3) implementation of log statements
• LOG INFRASTRUCTURE -  (1) log parsing, and (2) log storage.
• LOG ANALYSIS: : (1) anomaly detection, (2) security and privacy, (3) root cause analysis, (4) failure prediction, (5) quality assurance, (6) model inference and invariant mining, (7) reliability and dependability, and (8) log platforms
• Log Parsing - “textual similarity” between the log messages.
• Each log is converted to a binary vector, with each element representing whether the log contains that keyword
• Transformer - TEMPLATE2VEC (as an alternative to WORD2VEC) to represent extracted templates from logs and LSTMs to learn common sequences of log sequences

Root Cause Analysis

• By correlating log messages and resource consumption, their
• approach builds relationships between changes in resource consumption and application events.
• They propose a technique based on the correlation of console logs and resource usage information to link jobs with anomalous behavior and erroneous nodes.

Failure Prediction

• Utilize system logs to predict failures by mining recurring event sequences that are correlated

Paper #2 - Multi-Source Anomaly Detection in Distributed IT Systems

Key Notes

• Three categories-modalities: metrics, application logs, and distributed traces
• Word frequencies and metrics derived from the logs (e.g TF-IDF)
• Decompose the trace in its building blocks, the events/spans, and predict the next span in the sequence

Paper #3 -  LogBERT: Log Anomaly Detection via BERT

Key Notes

• LogBERT leverages the Transformer encoder to
• model log sequences and is trained by novel self-supervised tasks to capture the patterns of normal sequences. 

Baselines

• Principal Component Analysis (PCA) [19]. PCA builds counting matrix based on the frequency of log keys sequences and then reduces the original counting matrix into a low dimensional space to detect anomalous sequences
• One-Class SVM (OCSVM) [14]. One-Class SVM is a well-known one-class classification model and widely used for log anomaly detection [5,16] by only observing the normal data.
• IsolationForest (iForest) [7]. Isolation forest is an unsupervised learning algorithm for anomaly detection by representing features as tree structures.
• LogCluster [6]. LogCluster is a clustering based approach, where the anomalous log sequences are detected by comparing with the existing clusters.
• DeepLog [2]. DeepLog is a state-of-the-art log anomaly detection approach.
• DeepLog adopts recurrent neural network to capture patterns of normal log sequences and further identifies the anomalous log sequences based on the performance of log key predictions.
• LogAnomaly [23]. Log Anomaly is a deep learning-based anomaly detection approach and able to detect sequential and quantitative log anomalies.

Paper #4 - A Survey on Automated Log Analysis for Reliability Engineering

• Log event sequence: A sequence of log events recording system’s activities.
• Log event count vector: A feature vector recording the log events occurrence

Analysis Insights / Thoughts

• The query for selected values vs Bulk Upload of Data
• Usage patterns segmented for weekday/weekend / Trading hours
• Usage patterns across different time zones
• Usage patterns across different sections of applications
• Number of ad-hoc queries
• Restrict bulk upload to certain timezones / non-peak hours
• Two-stage commit - Upload and commit at a later stage

Big picture Notes

• Limit users to App Access during peak hours (5 calls during peak hours)
• Limit users to App Access during peak hours (10 calls during non-peak hours)
• Refer to replicated data in case of data that has stop-gap 5 hours delay
• Pagination of results
• Cache/reuse of results
• Identify maximum reported errors
• Patterns of errors over a weekday 
• User login activities and queries
• User value - Application usage vs Revenue
• User Action predictions 
• Take top 100 users, Plot the sequence of usage and see common flow/patterns

Diagnosis Perspective

• What is the blocking that happens between
• Page load query vs Search query
• Search query vs Data upload query
• Data upload vs Report download query
• Measure potential data conflicts that cause issues

Prediction Algorithms for User Actions

Execution model

• Understand problem statement
• Understand data sources
• Understand data access / permissions
• Frame NLP / Data / User level details
• Initial Analysis Scope
• Application Understanding
• Connects / Feedback

Diagnosis

• User based - Create APIs / Read APIs / Update APIs - Simple / Bulk / Delete APIs - Single / Bulk
• Do we track at UserId, Numberofcalls,Avgtime
• Nature of transactions - Realtime vs Reporting vs Bulk inserts vs Bulk Updates
• API calls across day by time
• %% Mix of workflow and common tables mapped / accessed by them - Time dimension added for pattern
• A,B,C @ Time T1
• A,B at Time T2

More Reads

• Detecting Anomalies in Software Execution Logs with Siamese Network
• How to Predict the Next Best Action to Progress Sales with TensorFlow 2
• Markov Chains in Python: Beginner Tutorial

Keep Thinking!!!

Contributions / Efforts / Building Teams

After years when we look back and see beyond the company, roles, projects, What is the impact or what we have learned or the teams we worked with, how do we remember it.

• Years you work vs Contributions
• Years you work vs Learning over time
• Years you work vs Mutual Win / Win projects
• Years you work vs Going beyond the comfort zone
• Years you work vs Self Realization
• Years you work vs Meaningful Work
• Years you work vs Influence over Peers
• Years you work vs Giving back to the community
• Years you work vs Sharing your learnings

Everything is part of building a high-performing team. Great teams come from trust, learning, picking the best ideas, and experimenting in the available time.

The only thing constant is finding/becoming a better version of myself compared to yesterday. Just in time answers vs forgotten memories vs Who you are vs What you learnt vs What you lost in the making is a never-ending quest to understand ourselves.

Keep Going!!!

Efforts / Learning / Skills / Dimensions

• To File a patent - Domain Knowledge + ML Knowledge + Uniqueness of Patent + Business Benefits
• To do a MVP - Study competitive products + Assess ML models + Pick relevant data + Get the ROI / Outcome
• To Scale a Model - Study Cloud / Deployment architectures + Scalability + Monitoring aspects 
• To Teach Something - Getting into specifics, Move from practitioner to researcher lens to get insights
• To get Ideas - Check blogs, newsletters, books

Everything counts to the dots of ideas, Everything connects in long run. Many things are not measurable with profiles but the impact is visible in the right team mix and focus. Coding after gathering required knowledge vs Searching for code and idea while you code both reflects your experience

Thinking is idea / logic / reading / listening. Coding is the method to connect those dots. A lot more things are there which we do not even think about when we write about our skills.

Keep Going!!!

Data Science Skills

Data Science Skills = Diversity Thinking = Database Skills + Domain Perspectives + Data Science thinking hat + Possibilities with available data

Bridging Birds Eye + Available Features + Expectations from customers is a refinement of picking the best of ideas, quick experimentation. Practitioner's perspective to apply multiple context/patterns is ongoing learning.

Myth of Expertise

• Big data - Current Trends
• NLP - Current Trends
• Vision - Current Trends
• GAN - Current Trends
• Transformers - Current Trends

Knowing vs Doing

• Prototypes in Current Trends
• Projects with Current Trends
• Go-Live with Current Trends

Knowing vs Doing vs Delivering vs Connecting the Dots is a different Skill. 

I look to solve a selected set of problems with four hats

• Database Developer/ BI perspectives
• Data Science Perspective
• ML Algos Perspectibes
• Customers Perspective what makes sense

It is a mix of all these thoughts that makes up a use case. One Skill does not make things happen. Unfortunately, these don't stand up in your profile but are even harder to measure relevant skills vs practical implications

Keep Thinking!!!

Tech Talk - Causal Inference in Data Science From Prediction to Causation

Key Notes

• Games Prediction - Higher Activity logins, Friends 
• It could be another way, They play games and make friends
• How do we increase the activity?
• Different segments of people - Games to Friends, Ask Friends and play games
• Observational metrics - Be mindful of hidden causes

• Measure versions of algos - A/B Testing
• Impact of Algo on different types of people
• Lower activity - Higher CTR
• CTR for different segments of users
• Segment people and see the behavior of each segment with experiments
• Combination of Experiments / Conversions / Measure of it
• ML Recommendations
• Split groups into different selections of the same category
• The choice for new Algos - Frequent Buyers
• Choice of old Algos - Low-frequency Buyers
• Purchase behavior trend over years
• Purchase behavior of new buyers
• Experiment - Conversions - Alerts (Forecast vs Actuals)

Frameworks

• Causal graphical models
• Potential outcome framework

• What would have happened if you did that?
• What would have happened if you had not done that?

Evaluate existing systems

• Old recommendations vs New Recommendation
• Measure forecast deviations against actuals qualitatively

Feedback loop informs current best algo!!!

Keep Thinking!!!