I build backend and AI-enabled systems that are secure, testable, and maintainable. My experience spans financial services, QA automation, technical education, and product engineering, backed by bachelor's and master's degrees in Computer Science.

• Backend systems — Built Java and Spring Boot services supporting 1M+ daily financial transactions, a 99.9% uptime target, and payment and core-banking integrations. Backend optimization improved transaction processing by 30%.
• Automation and reliability — Turned recurring validation work into repeatable automation, reducing manual workflow effort by 50% while improving regression coverage and release confidence.
• Applied AI and ML — Engineer agentic RAG, retrieval, memory, safety controls, and evaluation workflows with Python. My work also includes leakage-aware fraud modeling with causal features, chronological validation, hybrid XGBoost/FNN/LSTM inference, and drift monitoring. I focus on inspectable architectures, reproducible experiments, and measured failure modes—not demo-only model calls.

A multi-user inventory platform evolved from a PyQt and SQLite desktop application into a containerized React and FastAPI system.

Architectural highlight: Enforces tenant boundaries at the schema and service layers with organization-scoped constraints, OIDC/PKCE authentication, hierarchical roles, and audit events committed in the same PostgreSQL transaction as inventory mutations. Its legacy importer is idempotent, supports dry runs, and reports source-to-destination totals.

Python · FastAPI · React · TypeScript · PostgreSQL · Alembic · Keycloak · Docker Compose

A leakage-aware transaction risk platform that turns chronological card activity into authenticated fraud scoring and reviewer workflows.

Architectural highlight: Prevents temporal leakage with causal per-card features, immutable chronological partitions, and preprocessing fitted only on training rows. Its validation-selected log-odds fusion combines XGBoost, a feedforward neural network, and a causal LSTM, achieving 0.9767 average precision on the out-of-time holdout before deployment through an authenticated FastAPI runtime with drift monitoring.

Python · FastAPI · XGBoost · PyTorch · scikit-learn · Streamlit · Docker · Vercel

An educational medical-information assistant that routes symptom, severity, description, and precaution queries through specialized tools.

Architectural highlight: Combines a LangGraph ReAct orchestrator with per-session memory and a local TF-IDF/FAISS retrieval layer, keeping offline verification deterministic. A 60-case adversarial suite reports 97.67% Recall@k, 100% out-of-scope detection, and a 0% harmful-response rate.

Python · LangGraph · LangChain · FAISS · scikit-learn · Gradio · pytest · Docker

A merchant and payment-transaction REST API modernized from Java 8 to Java 21 and Spring Boot 3.5.

Architectural highlight: Treats Flyway migrations as the schema authority, validates the migrated schema through Hibernate, and runs tests against H2 in Oracle compatibility mode. The API uses resource-oriented routes, immutable records, RFC 7807 error responses, and decoupled slice tests.

Java 21 · Spring Boot 3.5 · Spring Data JPA · Hibernate · Oracle · Flyway · H2 · Maven

I'm always glad to connect with fellow developers, technology enthusiasts, and people learning their way through the field. Here's how you can reach me:

Email: meetyusufadamu@gmail.com

LinkedIn: linkedin.com/in/yusuf-adamu-9b3472115