Khaldoun - Details
Most people and most companies run on auto-pilot. This is a good thing because many processes just work. However, it is also true that many of us frequently feel we are losing control.
Khaldoun is a data science boutique that helps you side-step your or your organisation's auto-pilot.
When we work directly with a company, we use low-key data analytics & AI technologies to quickly identify & capture new potential. This doesn't require changing the entire organisation (or switching off the auto-pilot). Instead, we leverage key pressure points, which minimises dependence on internal ressources during and after implementation.
Aside from our open source products, our toolbox is very simple: power point, microsoft word, git, python, sql, docker. That's it. These tools allow us to prepare strategy documents, map processes, set up kpi & process cockpits, develop & deploy software solutions.
Typical initiative procedure
- 1) make status quo transparent: write a concept, conduct interviews, explore data, map processes
- 2) test out solutions: derive hypotheses, quickly iterate potential solutions, develop implementation plan
- 3) develop & deploy: deploy pilot solution, arrange handover
Areas with proven potential
- batch review of suspicious activity: many organisations have established review processes that are not rigorously monitored; therefore, we can identify new potential that has been missed so far.
- automation in core processes: many organisations forego investments into their processes for a long time and eventually decide for lift & shift; in contrast, we focus on identifying leverage points where automation creates non-linear results.
Why such a simple approach?
- There are many AI companies that develop fancy AI solutions. Many of these solutions & organisations are amazing. But most established organisations that hire these AI companies don't have the in-house competencies to deal with highly complex machine learning software. Usually, they don't even have the competencies to deal with their existing IT infrastructure. So they end up outsourcing a significant part of their core business without noticing it.
- We believe it is best to identify potential and realise it in an easy-to-understand manner. It is not as flashy and you might even leave some potential on the table. Yet, you'll maintain your independence and cut back the complexity.
Machine Learning innovation: Robotic agents
- Any legacy enterprise struggles with complicated, interrelated IT systems.
- Typically, it is not possible to deploy an algorithm within any of these legacy IT systems.
- Therefore, data scientists have two viable options to integrate:
- either they develop REST APIs that expose algorithms & convince the organisation to consume these APIs
- or their AI algorithms mimic a human specialists behaviour & act on their behalf.
- In most organisations, REST APIs need to conform to context-specific standards.
- Aside from understanding the business context, a data scientist needs to consult many other involved parties such as software engineers, IT architects and IT security specialists.
- Sometimes, the organisation is not even ready to deploy or consume standardised APIs.
- In contrast, if a human can grant an algorithm access to automate work for him, their is no integration work required because the algorithm can access the same as the human does (which is necessarily sufficient).
required tooling
- language learning: dashboard for learner & coach to see activity levels
- how many minutes per week did learner speak with AI
- what's AI-rated competence score this week
- learner-AI exercises (e.g. write story/essay together)
- core skills: CI/CD AI
- whenever work is ready for opening a PR, let AI critique work first & incorporate AI feedback
- basic setup: make sure everybody has a high productivity setup
- core setup (virtual envs, pre-commit hooks, linter, ...)
- habits for productivity (self-check-ins, quarterly OKRs, ...)
- basic training for git, ChatGPT, Github Copilot to integrate them in daily life
- idea school: engage with ideas that have potential to expand productivity
- socratic seminars on topics such as open exploration, structured execution, ...
Intended structure:
- Tunisian company employs Tunisian data scientists. German company contracts Tunisian company for remote work. European healthcare businesses contract the German company, which contracts the Tunisian company.
- Data scientists receive on-the-job guidance (videos, peer reviews, seminars).
- The European healthcare businesses can hire the Tunisian talent full-time over time, if they offer them a permanent position in Europe (work visa).
relevant skills
- 1) exploratory data analysis with sql & python
- 2) version control with git
- 3) let generative AI do your work
- 4) software design patterns with python
- 5) statistical foundations
- 6) non-linear thinking
- 7) local language proficiency
We train data scientists in Tunisia and they work remotely at European healthcare businesses.
- A well-educated Tunisian workforce has limited high-quality job opportunities. Europe is in need of enlarging its workforce.
Many non-Europeans that try are unable to secure a job in Europe because ...
- they don't close the language gap, which typically means having at least B1 in their target language,
- they collect certificates rather than build actual skill,
- they (or their agencies) do not have strong relationships with relevant recruiters in Europe,
- they are ill-prepared for the specific context they would work in.
Curriculum: We train candidates with a curriculum that is comprised of ...
- intense language learning aided by AI voicebots,
- intense software engineering & data science skill building aided by AI & peer reviews,
- actual work assignments from future employer,
- seminars on core topics to increase productivity (e.g. OKR-based work, deescalating tensions).
What does a candidate need?
- willingness to go all-in
- loves learning
- basic level of python (ideally also git, sql, bash)
- openness to become data science generalist (data science + engineering + strategy)
Why does it matter?
- Our education is expensive, time-consuming and has limited relevance in industry.
- However, simply educating anyone without specific relevance is a waste of ressources.
- Machine Learning has a veil of excellence that is largely unwarranted. While there are positions that in fact require cutting-edge competence & superior intelligence, most companies do not look for candidates with such a skillset.
- Instead, we can benefit from focussing on the essentials that make employees valuable.
Convictions
- We focus on staying small. We leverage productivity tools, esp. around AI.
- We focus on solving immediate problems with ambitious multi-year agendas.
Seminar on execution
- shake out your body, and get high energy from music
- explain Khaldoun & the trifecta of language, skills, execution and why we focus on execution today
- breath-holding exercise: show yourself you can do
things you think are impossible for you
- ask how long they think they can max. hold their breath
- explain the exercise and why they'll likely fail the first try
- do the exercise
- ask how group feels and what they noticed
- ask if anyone wants to share what he/she did that seemed impossible at first
- high energy state: motion creates emotion
- show Tunisian gesture and declare it enemy number one
- your environment wants to pull you down all the time (out of fear or unhappiness)
- it's your job to manage your energy in an estranged environment
- how to learn: total immersion, focus, model someone
- tell a story to visualise the message
- ask somebody how he would learn German, then refine it
- visualisation exercise: happy memories and memories you're going to create
- make an honest list of all the things endangering this future
- create a strategy how you'll have achieved your future