"I worked with Andriamarolahy R. for 1.5 years in a part-time capacity as a Junior Software Developer. During this time, he contributed to code implementation and revision in a methodologically complex innovation project, consistently demonstrating reliability and a solid grasp of best software engineering practices. His technical skills, particularly in Python (medical imaging algorithms, ML/AI, data visualization, data analysis, OOP), were at a mid-level when he joined, and he showed clear improvement over time. He was proactive in asking questions whenever clarification was needed, and he was always motivated and eager to learn. With his excellent work ethic and kind personality, he brought well-appreciated contributions to the project and team. I definitely recommend working with him."
Let's build useful systems.
I design and build intelligent automation systems that turn complex workflows into reliable tools. My work focuses on machine learning, computer vision, and data engineering to help organizations reduce repetitive work and unlock the value of their data.
I’m an AI Engineer focused on practical systems, from prototype to production.
From Problem to Solution
A deep dive into the AI systems and automation tools I've built from the ground up.
ECOWAS ID Card Classifier
In many organizations, people spend hours a day manually sorting scanned documents, often combing through thousands of pages to find and categorize ID cards. This...
Automated Tomato Ripeness Detector
Farmers lose revenue when tomatoes are picked too early or too late, and inspecting fields manually can take 2–3 hours per acre. That inconsistency translates...
PODLITE: Summarize & Unlock Insights from Podcasts
Listening to a 60-minute podcast typically takes an hour, meaning key insights are often missed or postponed. That’s a huge time cost for professionals who...
RAGi: Intelligent Document Chatbot
Searching a long manual or contract can take 30+ minutes, and generic chatbots often provide confident but incorrect answers (“hallucinations”) that lead to wasted follow-up...
Data Tables Extractor
Transcribing data from scanned tables typically takes 30–60 minutes per page and introduces manual errors that then require downstream cleanup. I built a Streamlit app...
In Their Own Words
I'm grateful for the relationships I've built and the trust I've earned along the way.
“