About Projects Experience Contact
Available for hire · Remote

OKORIE
CYNTHIA KALU

Backend Engineer building real systems

Python · FastAPI · PostgreSQL

I build backend systems that handle real-world problems — from secure APIs to AI-powered services with streaming, background tasks, and production-ready architecture. I ship software that works.

→ Currently open to backend roles (internship & junior) — remote or on-site

→ View Projects → Explore Live APIs ↗ → Download CV ↓
Live APIs → MediBook API ScholarIQ AI cynthia-kalu-o-portfolio.netlify.app ↗

01 — About

Building APIs that
solve real problems

I'm a backend engineer specialising in Python, FastAPI, and PostgreSQL. Over the past year, I've built and deployed backend systems that go beyond simple CRUD — integrating AI, handling concurrency, and managing real-world constraints.

Before tech, I spent 7 years in education and administration — working in high-pressure environments that required structure, communication, and problem-solving. That experience shapes how I design and build systems today.

My flagship project, MediBook API, is live in production — a healthcare booking system with AI features, background tasks, and OAuth. I enjoy building systems that solve real problems, especially where AI meets practical use.

Tech Stack

Python 3.13
FastAPI
PostgreSQL
SQLAlchemy 2.0
Alembic
JWT Auth
Google OAuth
OpenAI API
Google Gemini
Celery
Redis
Docker
Render
Git / GitHub
Pydantic v2
REST APIs
Async Python
Okorie Cynthia Kalu

Education & Certifications

ALX Backend Engineering 2025
PGD — Public Health ABSU
HND — Environmental Health Ebonyi

02 — Projects

Things I've
built and shipped

01
Featured · Live in Production

MediBook API

The problem: Healthcare booking systems lose patient trust when two people book the same slot simultaneously, or when AI recommendations are unreliable under load.

MediBook solves this at the system level — atomic slot-locking at the database layer prevents double-booking entirely, and AI urgency classification helps patients get the right slot, not just the next available one.

FastAPIPostgreSQL SQLAlchemy 2.0OpenAI GPT-4o-mini Celery + RedisResend Google OAuthRender
↗ Live Demo ↗ GitHub

What I built & learned

  • Eliminated race conditions — atomic slot-locking inside database transactions means two concurrent booking requests cannot both succeed, even under load
  • AI urgency classification — GPT-4o-mini reads patient symptoms and recommends slots by priority: low, medium, or high — improving appointment decisions, not just availability
  • Resilient AI integration — learned to design for failure: 15s timeouts, 2-retry logic with delay, and graceful degradation so booking works even when OpenAI is slow
  • Async background processing — Celery + Redis offloads email workflows (confirmation, cancellation, 24h reminders) so they never block a booking response
  • Secure multi-role access — JWT and Google OAuth with clinic and patient roles cleanly separated throughout the entire request lifecycle
02
Live · Production API

ScholarIQ AI

The problem: AI writing tools generate content confidently — including fake citations. This erodes trust in AI-generated academic work.

ScholarIQ adds a validation layer that detects hallucinated or invalid references, so the output can actually be trusted. It also adapts its writing style to the content type and academic level of the request.

FastAPI Python Google Gemini google-genai SDK Pydantic SlowAPI Docker Render
↗ Live Demo ↗ GitHub

What I built & learned

  • Citation hallucination detection — built a validation layer that flags suspicious, fabricated, or poorly structured references from LLM output before they reach the user
  • Context-aware content generation — learned to engineer prompts that adapt output by writing type, category, and academic level rather than returning generic text
  • Real-time streaming — implemented HTTP streaming for long-form responses so users see output progressively instead of waiting for full generation
  • Stateless API design — deliberate architectural choice: SQLite only for rate-limiting state, no heavy persistence, keeps the service fast and cheap to run
03
Live · Production API

Cerebro API

The problem: Most AI chat tools give you answers. They don't push back, challenge your reasoning, or help you think more clearly.

Cerebro is a real-time AI conversation API where the model is explicitly prompted to question assumptions and deepen thinking — not just generate. Built with full SSE streaming and persistent conversation history.

FastAPI Python PostgreSQL SQLAlchemy 2.0 async OpenAI GPT-4o-mini SSE Streaming Docker Render
↗ Live Demo ↗ GitHub

What I built & learned

  • Server-Sent Events streaming — learning to deliver AI output token-by-token in real time, so users see the response building rather than waiting for a full block
  • Socratic prompt engineering — the AI is instructed to challenge, question, and redirect rather than simply respond — a meaningful difference in system design
  • Stateful conversation persistence — full message history stored in PostgreSQL; users can return to any thread and continue where they left off
  • Fully async architecture — SQLAlchemy 2.0 async + asyncpg throughout; the deepest async system I have built so far

03 — Experience

Where I've
been and built

2024 — Present

Backend Engineer — Independent Projects

Self-directed · Remote

Built and deployed 3 backend systems using FastAPI, PostgreSQL, and AI integrations. Engaged publicly with senior engineers on X, implemented their feedback within 24 hours, and iterated rapidly on production systems.

2025

ALX Backend Engineering Programme

ALX Africa · Remote

Intensive software engineering programme covering Python, backend systems, API design, databases, and deployment. Graduated with hands-on project experience.

2021 — 2023

Freelance Web Developer (WordPress)

Freelance · Remote

Developed and deployed client websites including a mental health blog and an e-commerce platform. Worked directly with clients to translate requirements into working solutions, and maintained sites post-launch.

2020 — 2022

Remote Transcriptionist

Freelance · Remote

Delivered accurate transcription under tight deadlines, strengthening attention to detail, consistency, and remote work discipline.

2017 — 2024

School Administrator & Classroom Teacher

Various Institutions · Port Harcourt, Nigeria

Managed operations and coordination across teams in high-pressure environments over 7 years. Built strong communication, organisation, and problem-solving skills that now directly shape how I design and document backend systems.

04 — Contact

Open to opportunities

I'm currently open to backend engineering roles — internship or junior level, remote or on-site anywhere in the world. If you're building something interesting, I'd love to contribute.

✓ Message sent! I'll get back to you shortly.
✗ Something went wrong. Please email me directly at cynthiakalu1974@gmail.com