Project Inquiries

  1. The sheer volume of online information and products makes it increasingly difficult for users to discover relevant content or items that align with their preferences. This information overload often leads to decision fatigue and decreased user engagement. Recommendation engines address this problem by analyzing user data, preferences, and behaviors to provide personalized suggestions, guiding users towards items they are likely to find interesting or valuable. This not only enhances the user experience by saving time and effort but also significantly benefits businesses by increasing sales, engagement, and customer satisfaction.
  2. Play with the tool here! 🔛 (The VMs hosting the application take about a minute to start up on initial inference)
  3. https://recommend-ofiw.onrender.com/

    https://recommend-ofiw.onrender.com/

  4. Read about it here! 🔛 This tool was built using Plotly Dash.
  5. https://medium.com/@cwakhusama/understanding-recommendation-systems-in-e-commerce-633a336a70d2
  1. A customer support agent at an e-commerce company could use text summarization to generate a summary of a customer ticket that describes a problem with an order. The agent could then use NER to identify the key entities in the summary, such as the customer's name, order number, and the product they are having problems with. This information would allow the agent to quickly understand the issue and provide the customer with a resolution.
  2. Play with the tool here! 🔛 (The VMs hosting the application takes about a minute to start up on inital inference)
  3. https://ner-summary-headline.onrender.com/

    (NER Summerizer and Headline)

  4. Read about it here! 🔛 This tool was built using Spacy and a Flan & Bart LLM models.
  5. https://medium.com/@cwakhusama/spacy-for-named-entity-recognition-and-llms-for-text-summarization-and-headline-generation-74f35b15b35e
  1. Challenges of existing models.
❎Black-box nature of AI models. The complex internal workings of AI models make it difficult for doctors to understand how they arrive at their diagnoses.
❎Potential for bias and errors. AI models can inherit biases from their training data, leading to inaccurate diagnoses for certain patient populations.
❎Need for interpretability and explainabilty. To trust AI models in critical medical decisions, doctors need clear and concise explanations for their predictions.

Main advantage of XAI Solution.
▶️Feature importance analysis. XAI can identify the specific features in the patient's data that most influenced the model's prediction, providing valuable insights for medical professionals.
▶️Improves the medics knowledge for matching symptoms to correct diagnosis.

Benefits.
☑️Improved trust and transparency. XAI helps doctors understand the reasoning behind AI models, fostering trust in their predictions.
☑️Enables model to be audited by the Dr. on all the inferences made.
☑️Enhanced decision-making. By understanding the factors contributing to a diagnosis, doctors can make more informed and personalized treatment decisions.
☑️Reduced misdiagnosis. Models are trained with expert data and transfers learning to the user (medical practitioner)
  2. Play with the tool here! 🔛 (The VMs hosting the application takes about a minute to start up on inital inference)
  3. https://dem-8wfi.onrender.com/

    (DEM.V2 => Demp)

  1. Yes 😊.
  2. Consider a company that wants to predict demand for products and services or set a price discriminatory strategy based on historical data and factors such as day, area, time-of-day, and weather. This information could then be used to optimize inventory levels and pricing.
  3. Additionally, if the company wants to plan marketing campaigns. For example, retailers could target customers with special offers based on the products and services that they are most likely to be interested in, and the times of day when they are most likely to be shopping.
  4. Play with the tool here! 🔛 (The VMs hosting the application takes about a minute to start up on inital inference)
  5. https://psag.onrender.com/

    (Price discriminatory strategy)

  6. Read about it here! 🔛 This tool was built using Passive Agressive Algorithm.
  7. https://medium.com/@cwakhusama/online-ml-algorithms-a4b5c4b3425d
  1. Understanding which market segments your consumers belong in and their regional preferences allows us to tailor our product offerings and marketing strategies effectively. By doing so, we can better meet our customers’ needs and enhance their overall experience with our brand. But how can we notice this segments and other nested segments? Use graph analysis to Visualize your data to quickly pick out these segments and how they relate with the different regions, customers, products, shipping mode and state.
  2. Play with the tool here! 🔛 (This tool is hosted on Streamlit and migtht take a minute to run)
  3. https://appgraphnetworks-93pagujjuw25ejy5z5raxs.streamlit.app/
  4. Read about it here! 🔛 This tool was built using Pyvis.
  5. https://medium.com/@cwakhusama/graph-network-analysis-4920505ffc9a
  1. Assume you are a doctor operating in an urban set up where you are required to review a considerable number of patients daily. You employ a ML tool to assist with diagnosing the patients. In the process you are not sure about the criteria the model is using to make the diagnosis. You are more concerned about the accuracy of the model and whether it is biased in any way. To reduce the risk of Misdiagnosis and better understand the model predictions in order to improve trust in the ML tool, model explainabilty feature is integrated. Giving the Doctor the insights as to why the model arrived at a particular diagnosis.
  2. Play with the tool here! 🔛 (The VMs hosting the application takes about a minute to start up on inital inference)
  3. https://pragnosisexplainer.onrender.com/
  4. Read about it here! 🔛
  5. https://medium.com/@cwakhusama/model-explainabilty-xai-5e4b11822619

  1. Yes 😊.
  2. The need for county governments to optimize water distribution strategies can potentially improve the living standards of the residents, enable business growth, and reduce health related risks due to lack of water. Water utilities can use ML tools to optimize their water distribution networks by predicting the time of day and location where water demand is high. This can help to reduce water waste and improve efficiency. In areas where water is scarce, such tools can be used to predict the time of day and location where water shortages are likely to occur (Preventing water shortages). This information can be used to implement water conservation measures and prevent water crises.
  3. Play with the tool here! 🔛 This tool was built using Random Forest Classifier. (The VMs hosting the application takes about a minute to start up on inital inference)
  4. https://rmm.onrender.com/

  1. Yes 😊.
  2. Stable Diffusion can be used to generate product images, even for products that do not exist yet. This can be useful for e-commerce businesses that need to create product listings quickly and efficiently. For example, an online clothing retailer could use Stable Diffusion to generate images of new clothing designs, even before the clothes have been manufactured or instead of shooting different angles of a product with different backgrounds, you could simply have a few images and use Stable Diffusion to put it in different backgrounds and different contexts depending on where you are marketing or what product it is.
  3. Play with the tool here! 🔛 This tool was built using Stable Diffusion. (The VMs hosting the application takes about a minute to start up on inital inference)
  4. https://stablediffusionux.onrender.com/
  5. This model shares how to deliver a unique experience to users who login based on personal data collected (Rough idea of stable diffusion in production. User name: 123, Pass: 123)
  1. Well, i think Dash is one of your answers
  3. Integrating Data Analytics and Business Intelligence tools in legacy systems has never been easy with Plotly Dash. Consider a company that runs a bespoke system that is over 10 years old. How do we make the system relevant to the changes in industry that demand leveraging data to establish a competitive edge? I think the Answer to this is python’s Plotly Dash. Dash Is designed for Analytics and ML/AI applications.
  4. The link provides analysis of data for a local business. The objective is to understand how different metrics compare and their relationship to the KPIs.
  5. Play with the tool here! 🔛 (The VMs hosting the application takes about a minute to start up on inital inference)
  6. https://integrate-using-dash.onrender.com/ Password: root Username: root123
  7. Read about it here! 🔛
  8. https://dev.to/dallo7/why-do-i-like-dash-for-ml-and-ds-projects-bo4

  1. Yes 😊.
  2. 1. I have used Flask, Gradio, Streamlit and Dash to build ML Apps
  3. 2. Push code to Git
  4. 3. Created a CI/CD pipeline via a webhook on github to the deployment service
  5. 3. Deploy App to Aws or Azure or OCI or GCP or PythonEverywhere or Railways or Render or Huggingface
  6. 4. Inference
  7. Or
  8. You use Rest endpoints in your ML App to Inference the Hosted Model
  9. Or you can CI/CD using Jenkins webhook'd to Git
  • Yes 😊.
  • I can use Power BI, Tableu, OAS and Qlik View to create ETL pipelines from Live production DB to a Staging DB, from there create a data model and views then aggregate to a report
  • Most of the times i find myself using python scripts right after i use excel to validate and quickly clean and Transform the data. Pandas or Dask, Polars for preprocessing and Dash for Reporting does the job
  1. Yes 😊.
  2. Neo4j is a tool used to model graph data formats. To use it Download Neo4j from Neo4j.com and being writing some Cypher(Reach out if you get stuck!😉). Graph Network is mostly used to discover complex relationship. I like using Neo4j especially to model qualitative data for deep insights. Cypher Now and Then ... 😊👍
  3. For modelling graph data i can still use GNNs usually is when am integrating graph analysis to a visualization.
  1. Yes 😊.
  2. I have use YOLO v8 and v6 for Detection and Classification problems, I one of the projects i did in 2022 used Yolo V6 to detect students(21) faces coming in and out of school to determine how many students are in school at a target time. Used OpenCV to integrate model with the Camera and image analysis.
  1. Yes 😊.
  2. i Used stabilityai/stable-diffusion-xl-base-1.0 to create a swahili wrapper and Deployed the flask model on Render
  3. Play with the tool here!🔛 (The VMs hosting the application takes about a minute to start up on inital inference)
  4. https://sw-stablediffusion.onrender.com
  5. I have levraged huggingface models, Fine tuned Chituyi/opus-mt-en-sw-finetuned-en-to-sw swahili MT model and also finetuned an ASR engine for swahili transcriptions Chituyi/wav2vec2-large-xlsr-sw-sw300m-tr-colab. Deployed the translation model on Azure ML Studio and integrated the rest endpoints in an Application. Meanwhile, using the inference endpoints on Huggingface with Gradio UI running on my local machine to test the ASR model
  1. Yes 😊.
  2. So far i have written 11 concept papers 7 proposals 6 Projects Developed and 2 Reports. Some include, in the field of ML AND Health(Advancing Health in Marginalised Communities in Kenya)
  3. In the field of ML AND Education(Nlp to improve English pronounciations in Early Learners avaoiding Mother toungue effects)
  4. In the field of Computer Vision(Detecting and Determining number of students in school at a particular time)
  5. In the field of ML Forecasting and Prediction(Improving demand side water rationing through forecasting and prediction and Advicing discriminatory strategies on water pricing)

  1. Thank you for taking time to go through some of the skills I've acquired in 9+yrs. I look foward to working with you on some exciting projects 🤗🖥️📝