Welcome! Let's harness AI and Machine Learning to boost your business. Together, we'll create custom strategies to drive your business forward, making your vision a reality.
Brief Background
Hello! I'm Hyacinth Kwadwo Ampadu, an AI professional based in Ghana. With over 5 years of remote work experience serving clients globally, I specialize in delivering innovative AI solutions tailored to meet the unique needs of businesses. From healthcare to finance, I've successfully implemented projects spanning predictive modeling, time series analysis, Natural Language Processing, and Generative AI.
I offer expertise, reliability, and a passion for pushing the boundaries of AI. Let's collaborate to bring your company's AI initiatives to life.
SKILLS & SERVICES
Python Development
Cloud Computing
Data Science
Machine Learning
Engineering & Operations
Natural Language
Processing
Generative AI
Work Experience
May 2023 - July 2024
AI/Machine learning engineer (Remote) - WEBICIENT
​Besti (LLM powered Chatbot application):
-
Spearheaded backend and AI development, leveraging state-of-the-art language models in Python.
-
Implemented Langchain framework with Pinecone for efficient data storage and retrieval, ensuring data privacy with optimized metafilters.
-
Engineered a versatile web scraper extracting information from websites, documents, and cloud storage.
-
Employed FastAPI for web API development, integrated Pusher for real-time event streaming, and utilized GCP for deployment on Google Cloud Run, leveraging Docker for containerization and Grafana for real-time monitoring.
​Streamlining Investment Decision-Making with AI-Agents:
-
Led the development of an AI-driven project that reduced time and resources in researching companies for investments. The project automatically populates relevant information when new companies are added to Airtable, enhancing efficiency and accelerating decision-making.
-
Developed a web application leveraging AI agents and APIs to autonomously populate fields such as company details, competitors, and financial insights, streamlining data enrichment.
-
Employed FastAPI for agile API development, ensuring seamless integration of AI-driven data enrichment processes.
-
Deployed the solution on Google Cloud Platform using Cloud Run and Docker containers for scalability and reliability, fostering cross-functional collaboration for successful deployment and ongoing stability, delivering transformative outcomes in investment decision-making processes.
AI/Machine learning engineer (Remote) - Freelancer
Some Projects Completed for Companies:
August 2022 - February 2023
Enhancing Ghanaian Political Tweet Engagement:
-
Achieved a 27% perplexity decrease through unsupervised training(masked language modelling) using Distil RoBERTa to adapt to the Ghanaian political domain.
-
Leveraging the domain-adapted model, developed a predictive model for predicting tweet engagement.
-
Finetuned the GPT-3 model to optimize tweet virality by generating engaging variations for low-engagement tweets, ultimately yielding a 25% improvement in overall engagement metrics.
-
Strategically deployed models on Google Cloud Platform:
-
Utilized Cloud Run for serverless architecture.
-
Employed Docker containers for efficient deployment.
-
-
Collaborated cross-functionally for successful deployment and ensure ongoing stability.
July 2020 – May 2022
Forecasting NFT Trade Performance:
-
Worked with an Australian startup to develop and implement machine learning models for predicting future prices of NFTs.
-
Achieved 10% ROI on a modest NFT portfolio, validating the accuracy of predictions generated by our machine learning models.
-
Engineered a robust data pipeline using Spark, collecting and preprocessing data from diverse sources like blockchain, Parsec API, OpenSea API, and social media.
-
Applied advanced NLP techniques, including lemmatization, stopwords removal, and sentiment analysis (using TextBlob, Spacy, and BERT) for handling social media data.
-
Implemented diverse ML models such as Random Forest, Gradient Boosting, and AutoML for accurate predictions.
-
Conducted comprehensive feature engineering and selection to enhance model performance and interpretability.
-
Deployed the results from the model into a nice dashboard created using Tableau for easy viewing and understanding by the stakeholders.
January 2021 – January 2022
Optimizing Rail Safety:
-
Led the development of a predictive model for estimating rail temperatures in Jacksonville, achieving a 1- 1-degree Celsius error for predictions 48 hours into the future.
-
Utilized the XGboost model with historical and weather data, engineered key features across 25 locations, and optimized the model for enhanced accuracy.
-
Implemented data pipelines using Airflow, ensuring efficient data processing and model deployment in a production environment using Amazon web services(AWS).
-
Collaborated cross-functionally to integrate the model into the company's platform, delivering valuable insights and contributing to ongoing improvements.
-
Successfully aided railway safety officers in planning safety measures through precise rail temperature predictions.
May 2020 – June 2021
Published Insights in COVID-19 Prediction:
-
Collaborated with a Dubai-based client to develop a highly accurate and interpretable deep learning model for predicting COVID-19 mortality rates and ICU-admission likelihood in time-bound and resource-limited scenarios.
-
Published results and the model-building process in a high-impact medical journal, showcasing its potential to aid frontline doctors in classifying patients in time-bound and resource-limited scenarios. •
-
Executed extensive data pre-processing and cleaning, employing diverse techniques to handle missing and unbalanced data, ensuring the model's robustness.
-
Constructed the model using PyTorch and fine-tuned it using advanced techniques like early stopping, learning rate scheduling, and hyperparameter experimentation, achieving high accuracy and generalization
-
Analyzed the model's behavior, identifying influential features contributing to predictions, enabling doctors to comprehend underlying factors leading to patients' ICU-admission and mortality risks.
June 2019 – February 2020
Optimizing Shipping Costs:
-
Collaborated with an Indian consulting firm to devise a machine learning solution for predicting shipping rates across cities and countries in Europe, enhancing customer experience.
-
Developed and implemented diverse ML models, including regression analysis, decision trees, and random forests, to accurately predict shipping rates.
-
Worked with a team of data scientists and engineers to preprocess, analyze, and visualize extensive datasets containing information on past orders and shipping rates.
-
Collaborated with cross-functional teams to deploy the ML model using MLOps best practices, such as containerization and continuous integration and deployment (CI/CD).
-
Iteratively improved model performance based on user feedback and fresh data, ensuring ongoing optimization of shipping cost predictions.
Personal Projects
Project Description:
With the rise of social media use among younger generations, it is important to ensure that the content they are exposed to is appropriate and free from harmful language.
This project builds a web application designed to promote safe online social interactions for children. The app flags these tweets, alerting the user of the potentially harmful content, and then offers recommendations on how to rephrase their tweets that are more kid-friendly or appropriate for children.
​
Code:
https://github.com/JoAmps/KidFriendlySocial
​
Blog:
​
​
Project Description:
Lead conversion is a crucial aspect of business, especially for app and website companies. Identifying and predicting which leads are likely to become paying customers is important since those individuals contribute a lot to the company's profit margin. This project deals with predicting these leads so that the required and appropriate arrangements are made to enable them to complete the process of converting, and to keep making profits for the company and business.
​
Code:
https://github.com/JoAmps/Lead_conversion_prediction_for_a_mobile_app_company
Blog:
​
Project Description:
Predicting and identifying the sentiment of users on an app is key for businesses and app creators to identify what the customers really feel and think about the app and make the necessary changes to address their needs to keep them from churning and using other fitness apps, so as to keep making a profit. This project deals with building models to predict sentiments of reviews from users and deploying it on the cloud, on the various fitness apps so businesses can make informed decisions going forward
​
Code:
https://github.com/JoAmps/bert-based-health-and-Fitness-sentiment-system
Customer churn prediction in a Vehicle Insurance company In Ghana
Project Description:
Customer churn prediction is an important business activity for any company. The ability to predict this is very key so that companies identify such individuals and make the necessary arrangements to keep them since the company thrives on getting customers subscribed to their insurance. This project deals with solving this important problem in this company, to help the company keep making a profit.
​
Code:
https://github.com/JoAmps/Churn-prediction-in-a-vehicle-insurance-company-in-Ghana
Project Description:
The ability to forecast the prices of any commodity is crucial for companies, traders, and individuals, to know the price of their commodity in the future, so they make better plans using this knowledge. NFTs are the new craze that everyone is jumping on, there's potential money in it, and this project deals with forecasting these prices, to know if the NFT price would grow or decline in the future, so individuals can know whether to buy this nft in the particular week or sell it if they already own one
​​
Code:
https://github.com/JoAmps/Forecasting-average-weekly-prices-of-the-Parallel-Alpha-nft
Blogposts
Here are some of my Blog posts, click on the respective images to read!
Unlocking ML Opportunities: Building a Strong Portfolio for Self-Taught Career Launch- PART 1
I wrote about the approach to building quality portfolio projects to launch your career in ML
Transforming Leads into Customers: Optimizing Lead Conversion with Predictive Modeling
I wrote about the approach to optimizing lead conversion utilizing predictive models
MLOps Best Practices for Machine Learning Applications: A Case Study of KidFriendlySocial
I wrote about the approach to building machine learning systems following the MLOPs best practices
My Experience in the Udacity Machine learning DevOps(MLOPS) nano degree
I wrote about my experience, and the courses and projects I built in this nano degree
Dropout in Deep learning
Understanding Dropouts in Deep Learning to reduce overfitting
Random Forests Understanding
Intuition and Implementation on a key algorithm to reduce overfitting in tree based algorithms
Decision Trees
I wrote about the intuition and implementation of the first tree-based algorithm in machine learning
Understanding of Support Vector Machine (SVM)
Explanation of the support vector machine algorithm, the types, and how it works
Normalization in Deep learning
I wrote about the different types of Normalization in Deep Learning. A very useful technique to avoid overfitting and generalize your model better.
Yolov3 and Yolov4 in Object Detection
Explanation of object detection with various use cases and algorithms. Specifically, how the yolov3 and yolov4 architectures are structured, and how they perform object detection
Linear and Logistic Regression
I wrote about the intuition and implementation behind the base algorithms for supervised machine learning
Research Publications
My research papers that have been published
Interpretable Tabular Deep Learning for the prediction of ICU admission likelihood and mortality of COVID-19 patients
The global healthcare system is being overburdened by an increasing number of COVID-19 patients. Physicians are having difficulty allocating resources and focusing their attention on high-risk patients, partly due to the difficulty in identifying high-risk patients early. COVID-19 hospitalizations require specialized treatment capabilities and can cause a burden on healthcare resources. Estimating future hospitalization of COVID-19 patients is, therefore, crucial to saving lives, this is what our publication addresses
Click on image to be redirected to the paper
