Notes from 24/8–31/8

AI

fast.ai’s revamp

Need to pick this up sometime

The Most In-Demand Tech Skills for Machine Learning Engineers

• Data engineers are responsible for making data pipelines. They create systems that ingest data, store it, and transform it into a useable form for data scientists and data analysts. 💻
• Data scientists create machine learning models to help drive decisions and create predictive tools. They might also do the experimentation and statistical inference used to draw causal conclusions. This second role is more of a traditional statistician’s role. 🧪
• Data analysts find insights in existing data and share that information with stakeholders. 📊
• Machine learning engineers take the proof-of-concept machine learning models data scientists create and turn them into scalable, optimized, serve-able models for use in APIs and apps. 👩‍💻

Blockchain

The National Bank of Cambodia boosts financial inclusion with Hyperledger Iroha

The NBC was looking for a blockchain system that could provide:

• A permissioned network: For security, the central bank wanted to approve every entity on the system, with each permission having separate protocols to control the level of access.
• Confirmed transaction finality: To protect the integrity of all accounts, once a transaction was validated, it had to remain immutable.
• Byzantine Fault-Tolerance (BFT): To ensure the continuity of the system, even if one or more nodes were compromised or offline, the system needed this approach to reach consensus.
• An account management system: This would balance all accounts in the ledger with each transaction, so that the bank could easily supervise transactions.
• Simple architecture: This would give a smaller attack surface with fewer vulnerabilities to make the system easier to protect and maintain.

The process works like this:

• With a smartphone and an ID, a citizen proves their identity by taking a document photo and a selfie that are verified by a sophisticated AI algorithm
• The user associates themselves with one of the Bakong member banks
• From then on, they can use Bakong to buy and sell from participating merchants, or to send money to friends or family through P2P transfers

For individuals, Bakong serves as an e-wallet, enabling mobile payments and online banking. With a personal QR code, users can transfer money without sharing any personal information.

For merchants, the system processes cashless transactions with customers, other merchants, and their own bank. Merchants that accept Bakong can also be viewed in a map inside the app, and more than 500 merchants are now registered.

For banks, Bakong reduces the high fees for interbank transfers.

And for everyone, the system is reliable and fast. Transactions take less than 5 seconds, with system throughput exceeding 2,000 transactions per second. This is faster and more robust than the older system it replaced.

Data

Know What Employers are expecting for a Data Scientist Role in-2020

This extremely meta blog post

Product

Observing the Double Diamond process in practice

The Guide to Product Metrics

Design Thinking to Design Doing

Architecture Playbook

I suppose it’s good to familiarise yourself with some of these

New to Product Management? Here are 6 foundations you should build!

1. Simplify problems by asking uestions

Ask the following questions:

1. Can a layperson understand the solution you’ve proposed?
   A useful question that forces you to empathize with someone who doesn’t have the same system/domain understanding as you.
2. Does this product require a manual?
   If your app/product requires a manual to navigate, you’ve done something wrong. Interfaces should be intuitive.
3. Does this feature reduce complexity? Or adds to it?
   Your feature/solution should reduce complexity/chaos in the system. Uploading an image shouldn’t require 3 levels of confirmation. It only adds to the complexity of the image uploading experience.
4. Is the solution moving the complexity from the customer to the technology? If yes, how reliable is the technology?
   Ideally, anything that can be done by technology should be done by technology. You need not add the onus on the user! However, the technology behind it should be reliable. Asking the development team the reliability of the solution helps in figuring out if the user will have a smooth experience.

2. Balance customer needs with business viability

This does not mean you won’t work on the costly feature, but your prioritization matrix should account for the cost to build. Note that the cost here denotes both money and time

3. Communicate according to your audience

Subtle, concise, and correct are the 3 adjectives stakeholders should use to describe your communication. Where a graph would do, don’t send a paragraph.

4. Good is better than perfect

5. Set consistent product and feature

This is so you can make fair comparisons

1. Define & share feature goals with stakeholders & development team

2. Announce releases across the board in your organization

3. Sprint demos — even after a particularly hard sprint

4. Bi-weekly updates to your stakeholders on product progress