🟢 Open to work
I build Python and NLP solutions for automating document and workflow tasks, learned through hands-on externship projects.
Portfolio of externship work and projects showing Python code, document processing, and NLP solutions aimed at automating business workflows.
Outamation · Aug 2025 · ✅ Verified by Extern
Curious about how AI is changing the way businesses operate? Dive into real-world AI projects at Outamation, mastering Python, NLP, and document processing. Gain hands-on skills in automating complex workflows and extracting powerful insights, preparing you for success in tech-driven industries. Top-performing externs may get the opportunity to interview with the host company for an internship!
Analyzed 90,000+ health records using Pandas, NumPy, and scikit-learn to train an XGBoost classifier achieving 81% ROC-AUC. Built a React + FastAPI application with SHAP visualizations to explain how indicators influence risk predictions. Deployed the full-stack application on Vercel + Render, displaying outputs into clear interpretations for users.
Engineered multi-channel skeleton features (position, velocity, bone-vectors) to improve model input representation. Iterated through CNN, TCN, and BiGRU-Attention architectures, raising validation accuracy from 55% to 74%. Achieved 74% and 84% accuracy from the final model under Cross-Subject and Cross-View evaluation benchmarks.
I build Python and NLP solutions for automating document and workflow tasks, learned through hands-on externship projects.
I am Benjamin Alimele, a Computer Science student focusing on Artificial Intelligence. I work as a Technical Assistant at Acadia Launchbox and as an Undergraduate Research Assistant at Acadia University. I evaluate AI models for Outlier AI, have experience in workflow automation from an externship, and build full-stack ML projects using Python, React, FastAPI, and modern ML libraries.
Externships
Outamation: AI-Powered Workflow Automation Externship
Completed · Aug 2025
Experience
Technical Assistant
Acadia Launchbox · September 2026 – Present
Undergraduate Research Assistant
Acadia University · June 2026 – Present
AI Model Evaluator
Outlier AI · July 2025 – Present
Resident Assistant
Acadia University · Aug. 2026 – Present
Workflow Automation Extern
Extern · May 2025 – Aug. 2025
Education
Acadia University
Bachelor of Science in Computer Science with Honors (Co-op) | GPA: 4.19/4.33 | Dean’s List · Class of 2028
Skills
✅ Verified by Extern
Curious about how AI is changing the way businesses operate? Dive into real-world AI projects at Outamation, mastering Python, NLP, and document processing. Gain hands-on skills in automating complex workflows and extracting powerful insights, preparing you for success in tech-driven industries. Top-performing externs may get the opportunity to interview with the host company for an internship!
The work examined how LLMs, OCR, and Python-based pipelines process mortgage documents and produce machine-readable data. Tasks included explaining LLM behaviour, cleaning and denoising document data in Colab, extracting fields and bounding boxes from PDFs, building a hybrid RAG pipeline with LlamaIndex, and benchmarking open-source models. Notebooks, code, and experiment notes were submitted.
I produced runnable Colab notebooks and code that cleaned mortgage data, denoised document images, extracted fields and bounding boxes from PDFs, implemented a hybrid RAG pipeline with LlamaIndex, and compared multiple open-source models in that pipeline.
Analyzed 90,000+ health records using Pandas, NumPy, and scikit-learn to train an XGBoost classifier achieving 81% ROC-AUC. Built a React + FastAPI application with SHAP visualizations to explain how indicators influence risk predictions. Deployed the full-stack application on Vercel + Render, displaying outputs into clear interpretations for users.
Engineered multi-channel skeleton features (position, velocity, bone-vectors) to improve model input representation. Iterated through CNN, TCN, and BiGRU-Attention architectures, raising validation accuracy from 55% to 74%. Achieved 74% and 84% accuracy from the final model under Cross-Subject and Cross-View evaluation benchmarks.
