Olusegun Ajose

• Working as part of a remote team to get buy-in from business owners and navigate AI product decisions around resource allocation, projects worth developing, prioritization, trade-offs worth making, and metrics to evaluate. • Designed a Retrieval Augmented Generation (RAG) based product, reducing idea-to-production time by 90% and improving AI adoption rate by 80% across UNESCO sectors. • Developed an AI-powered report system that enabled UNESCO sectors to synthesize information from member states reports such as Global Education Monitoring and Committee Activity Reports – reduced reporting time from 3 weeks to a few hours. • Engineering an AI-driven system to reduce search time and increase accessibility of UNESCO's 70-year audio archive (with 50,000+ hours of multilingual content). Tech Stack: Azure (Data Factory, VM, PowerBI, Databricks), Python, FastAPI, LLMs, LangChain, GPT, Mistral, Ollama, Prompt Engineering, AI Agents, Gitlab, Chainlit

• Collaborated with 6 clients to translate business needs into scalable data warehouse architectures for business intelligence. • Architected and implemented data architecture for high-volume analytical data, managing 1TB+ of streaming data and CDC scenarios, benchmarked using TPCH datasets for performance optimization. • Built a real-time data warehouse using Apache Airflow, PostgreSQL, GCS, Iceberg, Kafka, and Superset, enabling real-time ingestion, processing, and visualization of over 10 million transactions for a retail client. • Led the development, shipping and monitoring of an Agentic System that streamlined data querying, modeling, and management—enhancing decision-making efficiency and reducing manual data handling overhead. Tech Stack: GCP (Bigquery, Dataform, Cloud SQL), Python, SQL, Iceberg, Datafusion, Debezium, Celery, Kafka, Agno, PyArrow

Working within the Digital Factory team, I led the implementation of AI solutions, utilizing Large Language Models like GPT-4, Llama 2, Flan-T5, and transformers library to streamline business processes. My journey has included: 🔹 Developed an LLM-powered Slack chatbot (pulling data from different knowledge bases) that improved technical support by automating 85% of technical inquiries and reducing response wait time by 70%, while also fixing documentation gaps. 🔹 Engineered a web analytics dashboard and reporting system to analyze technology stacks and security vulnerabilities across 15,000+ URLs, resulting in 40% faster decision-making for web infrastructure migration. 🔹 Used PEFT and QLora techniques to fine-tune and adapt Llama 2 model to Danone’s Nutritional team specific patterns, vocabulary, and writing styles, resulting in more natural and coherent language generation within the Nutritional domain. 🔹 Development Practices: I established a robust CI/CD pipeline using GitHub Actions and Docker, integrated with Amazon ECR while implementing Single Sign-On (SSO) access for enhanced security. Tech Stack: AWS (ECR, ECS, Lambda, S3, Sagemaker), Azure Open AI, Power BI, Python, SQL, Numpy, Pandas, LangChain, Slack Bolt, Huggingface, Streamlit, Github, Docker, Airflow, GitHub/GitHub Actions, Confluence

🔹 AI PoC Collaboration with Design to Cost Deployment: In close partnership with the Design to Cost Deployment Director, I played a key role in developing an Artificial Intelligence Proof of Concept (PoC) that showcased the potential of AI in addressing complex product data challenges. 🔹 Data Infrastructure with Dataiku: Developed a robust data infrastructure using Dataiku, enhancing the quality, accessibility, and availability of critical data. This facilitated smoother data-driven processes and expedited decision-making. 🔹 Developed NLP-based feature extraction system for unstructured train technical documentation, automating data capture of 50+ technical attributes including safety systems, engine and propulsion, onboard technology, etc. 🔹 Leveraged HuggingFace Models to develop a topic classification system with a taxonomy of 500+ categories, enabling the processing of several documents per day. Tech Stack: Dataiku, Python, Numpy, Pandas, Huggingface

🔹 Empowering Healthcare with AI Chatbot: Fine-tuned the Flan-T5 model using a medical QnA dataset to create a medical chatbot. This AI-powered solution seamlessly assists users in seeking accurate and reliable medical information. 🔹 Quality Control Application with Deep Learning: Developed and deployed a Quality Control application that detects and flags manufacturing defects (such as voids) in electronic components or circuits. Streamlit, Docker, AWS Elastic Beanstalk, Yolov8, Roboflow, and Meta SAM model 🔹 Beehive Health Monitoring: Orchestrated real-time data streaming from IoT (Raspberry Pi) to AWS Kinesis to power the continuous monitoring of beehive health, harnessing live data insights to ensure the well-being of these crucial pollinators.

With a unique mix of strong business acumen and software development domain expertise, I worked with new and established companies to integrate machine learning modela with web applications. I led the successful integration and deployment of an internal web application enhanced with AI functionalities to predict the ability of loan applicants to repay the loans. By synergizing my expertise in application development with data science techniques, I orchestrated a solution that offered advanced features and insights to end-users. • Developed and integrated an ML-powered loan repayment prediction model into LoanWise, enabling the loan approval team to cut decision time from 7 days to 3 days. • Developed ATMPlus to provide real-time updates and error alerts for 5,000+ ATMs across 5 commercial banks – improving uptime.

Joining the innovative Personal Banking Cloud Division team at Appzone Group, I supported on-premises solutions for leading commercial Banks in Nigeria. Some of these applications automate banking operations, interconnect financial institutions, and provide end-users with access to a broad range of banking and payment services. As part of the Critical Response Team investigating and resolving application issues escalated by clients’ IT teams, I identified the root cause of application issues including database, code, operating system, web-server, integration, networking, and hardware issues. Focusing on clients’ experience, I leveraged a system-based approach to ensure that defects/bugs are resolved in compliance with the SLA while preventing future occurrences. Always up for a challenge, I developed enhancements as requested by clients, unit tested and deployed these features into production environment after passing User Acceptance Test (UAT). I integrated our software solutions with the banks existing systems as well as third-party providers like Financle, Pajuno, and Zoho CRM. In this client-facing role, I utilized well-honed communication skills and deep product expertise to educate clients on the journey – explaining the “whys” and significant value of our products. Tasked with ensuring proper handover, deployment, and integration of applications, I ensured all required technical details of the application are documented in the technical configuration document template. I also provided subject-matter expertise, customer advocacy, and analysis through all phases of the system integration lifecycle. ⚡ Technologies: ASP.NET, MVC, C#, .Net, Web API, CSS, JavaScript, Zoho CRM, Visual Studio, NHibernate, and SQL Server