FLEXIBLE
FUNCTIONS

An industrial research and development AI lab building practical use cases using current state of the art machine learning systems in different domains.

Our Vision

Deliver value with AI through practical use cases

AI has been called the new electricity and has the potential to transform multiple industries. We believe we are entering one of the greatest eras of technology.

While we hear a lot about potential use cases, much of the value in AI is still untapped and undiscovered. To unleash the potential of AI, we are bridging the gap between research and real-world applications by building the application layer for AI.

Our Mission

Build components of the AI application layer.

Our Approach: Learning By Doing

Flexible Functions is an industrial research and development AI lab building practical use cases using current state of the art machine learning systems in different domains. We do this while researching how to build new user products and experiences and integrating foundation models.

We expect the practical aspect of building use cases for AI will inform research on how to take advantage of AI to build new products and experiences. This will also inform us on the challenges that currently prevent AI from being more prevalent across various industries.

We believe we can only build useful applications by engaging end users, domain experts to understand the problems they are facing and building custom solutions for them. This approach will increase the adoption of AI and in turn lead to increased investment in scaling the field.

We believe useful AI applications emerge when we:

  • Engage with end users and domain experts
  • Identify their most pressing challenges
  • Apply AI solutions to these specific problems
  • Iterate based on real-world feedback
  • Scale successes to similar use cases

The Problem

  • AI is called “the new electricity” with transformative potential
  • Despite countless theoretical use cases, practical implementation lags behind
  • Most AI value remains untapped and undiscovered
  • There's a significant gap between research and real-world applications

What we are doing

1

Open source

We believe in sharing knowledge and contributing to the AI community.

2

Build rapid ML pipelines

Creating efficient pipelines for fast experimentation and iteration.

3

Leverage foundational models

Leveraging the power of the latest foundation models to build new products.

4

Develop open GUI low-code workflow tools

Making AI easy to use for non-technical domain experts.

5

Publish technical examples and guidance

Sharing our learnings and best practices with the community.

6

Benchmark with Kaggle competitions

Testing our approaches against industry competition.

7

Identify and build new use cases

Continuously exploring new applications of AI technology.

Research Focus

Key Questions

  • How can we make AI's predictive power useful in practice on a larger scale?
  • How can we use foundation models to improve existing products or create new experiences?
  • How can we make AI more efficient and cost-effective making it easier to apply in practice?
  • How do we connect model results to business decisions?
  • What metrics truly align with business outcomes?
  • Which novel use cases have the highest ROI potential?

Interesting Research Directions

  • Multimodal models
  • Small-data machine learning
  • Unsupervised and reinforcement learning
  • Synthetic data generation for edge cases
  • Low-code workflow tools for rapid experimentation
  • Better / finetuned prompting
  • Applying AI to the Enterprise, Healthcare, Creative work (Branding, design) and Agricultural sectors

Use Cases

Use Case 1: Intelligent Demand Forecasting

The Challenge:

  • Supply chains are increasingly vulnerable to disruptions
  • Overstocking leads to waste, tied-up capital, and missed opportunities
  • Understocking creates fulfillment issues and customer dissatisfaction

Current state of the art:

Traditional time series analysis like ARIMA, statistical methods

Our Solution:

We leverage an enterprise's various data sources to build a dynamic inventory management system. This uses an ensemble of advanced ML algorithms such as large-scale neural networks, Automl solutions, and the more traditional gradient boosting machines to predict with pinpoint accuracy:

  • What every customer will buy
  • When they wll make each purchase
  • Which location they will buy from
  • What price they are willing to pay
Our system can forecast demand at all levels with high accuracy, predict when and what customers will purchase, optimize pricing and customer experience, and ensure the right products are in the right place at the right time. This is the future we're building at Flexible Functions AI.
Demand Forecasting Demo

Use Case 2: Assisted Differential Diagnosis

The Challenge:

  • Medical diagnosis is complex and time-intensive
  • Healthcare providers in resource-constrained settings face significant challenges
  • Diagnostic errors can lead to improper treatment and poor patient outcomes

Current state of the art:

Coming up with differential diagnoses is currently done by humans based on signs and symptom presentations, The Doctor’s knowledge, and clinical guidelines.

Our Solution:

An advanced LLM application built on user inputs, disease labels, and the Uganda clinical guidelines. Once patient complaints, medical history, etc are passed to it, it returns a differential diagnosis. We are building this to assist doctors.

We are going for this due to our technical expertise given that Daniel Hosana is a doctor, and as a coder Silver can execute on it.
Diagnosis Assistant Demo

Roadmap

Phase 1

Launch initial use cases

Launch initial use cases in inventory management and healthcare

Phase 2

Expand our toolkit

Expand our toolkit with open-source workflow solutions

Phase 3

Build an ecosystem

Build an ecosystem of industry-specific AI applications

Phase 4

Partner with domain experts

Partner with domain experts to penetrate new markets

Phase 5

Develop a comprehensive platform

Develop a comprehensive AI application layer platform

Company Values

Solve hard problems

We break down complex challenges into small solvable problems.

DIY approach

We believe in hands-on work and taking ownership of challenges.

Open source support

We contribute to and benefit from the open source community.

Learning through doing

We believe in practical experience as the most effective path to mastery.

Clear communication

Explicit, clear communication without assumptions (Explicit over implicit).

Data-driven decisions

We make decisions based on data and evidence rather than intuition.

Founding Team

Daniel Hosana

Daniel Hosana

Research Scientist

Medical doctor with coding experience

Silver Rubanza

Silver Rubanza

Chief Technology Officer

Machine learning engineer with experience building end to end machine learning solutions, sales and marketing experience.

Wilson Ssukwe

Wilson Ssukwe

Operations Officer

Experienced in running business operations at scale