Abel Addis · Backend engineer, AI systems · Ethiopia (UTC+3)I build the systems behind AI products.

Typed APIs, async workers, data pipelines and retrieval you can trust, shipped to production with FastAPI, Django and PostgreSQL.

See the work Get in touch

AndelaBackend Engineer, 2025–26
AWS CertifiedCloud Practitioner
BSc Software EngineeringArba Minch University, 2026
2+ yearssoftware and product work

I care less about adding AI for its own sake, and more about making it useful, reliable and connected to real product workflows.

Abel, after moving from UI/UX and full-stack work to the backend

How a request moves through the systems I build.

Most AI features share one shape. A request comes in, the backend finds the right data, the model reasons over it, and a grounded answer goes back. Scroll to follow one.

POST/v1/askreceiving

01 / 06

User or system

A person or another service sends a request: a question, a search, or an event to process.

Web, mobile or service-to-service

Example payload stage 1 of 6

POST /v1/ask
{
  "user": "u_42",
  "q": "Which plan fits a 5-person team?"
}

1User
2API
3Database
4Knowledge
5AI
6Response

Userapp or service

API endpointFastAPI

Job queueCelery + Redis

PostgreSQLrelational

Vector DBpgvector, Qdrant

KnowledgeRAG context

AI modelany LLM

Response200 OK

01
User or system
A person or another service sends a request: a question, a search, or an event to process.
Web, mobile or service-to-service
02
API endpoint
A typed endpoint validates the input, checks auth and rate limits, then routes the work. Slow jobs go to a background queue.
FastAPI, Pydantic, Celery
03
Database
Two lookups run side by side: exact records from PostgreSQL, and meaning-based matches from a vector index.
PostgreSQL, pgvector, Qdrant
04
Knowledge
Results are ranked and trimmed into grounded context: the right facts, each with a source the answer can cite.
RAG, ranking, citations
05
AI model
The model reasons over that context only and returns structured output. Providers can change without touching the app.
OpenAI, Gemini, DeepSeek, Groq
06
Response
The user gets an answer grounded in real data, with sources they can check, streamed back to the interface.
JSON, streamed to the UI

Selected work

Four products, each starting from a question someone actually has.

Insight Stream

AI-powered content intelligence

Ever scrolled through hundreds of articles and Reddit threads just to find the few that matter to your work?

FastAPI
Celery
pgvector
LLMs

Open project AI and async pipeline

In development

MoodVerse

AI-powered spiritual reflection

Ever known exactly how you feel, but not which words to search for to find scripture that speaks to it?

FastAPI
Gemini
pgvector
React Native

Open project AI, retrieval, mobile

In development

SEO Maker

AI-powered SEO intelligence

Ever wondered why a site isn’t getting the search visibility it deserves, when the problems are buried across hundreds of pages?

Crawlee
BullMQ
LLMs
Next.js

Open project AI, crawling, queues

Adding an AI trading assistant

AlertChain

Automated crypto price alerts

Ever waited for a coin to hit a price, then missed the moment because you weren’t watching the chart?

FastAPI
Celery
Redis
PostgreSQL

Open project Scheduled jobs, notifications

That’s the highlight reel.

There’s one more in the full list. See all five projects side by side, filter them, and open the write-ups.

All projects

Where I’ve shipped.

Each role pulled me further toward the backend, from wireframes to merchant dashboards to production AI services.

Full career log

2026, currentCurrent
Freelance engineer
Backend and AI systems
Building and shipping backend and AI-focused projects for clients and portfolio work.
- Backend
- AI integration
- Client projects
Now
Oct 2025 – Jun 2026
Backend Engineer
Andela
Shipped production FastAPI services with AI-integrated features, using Celery and Redis for heavy jobs.
- FastAPI
- LLM APIs
- Celery
- Redis
2025–26
Feb 2025 – Aug 2025
Full-Stack Developer Intern
Santim Pay
Built merchant-dashboard features with Vue.js, Nuxt, Pinia, FastAPI and PostgreSQL.
- Vue.js
- Nuxt
- FastAPI
- PostgreSQL
2025
Oct 2023 – Feb 2024
UI/UX Designer Intern
Rabit Delta
Wireframes and usability testing that shaped product and feature decisions.
- Wireframes
- Usability testing
2023

How I build.

Six rules I keep coming back to, and where each one shows up in my own projects.

01 / 06

Ship, don’t just prototype

I prefer working features that survive real requests, failures and changing requirements.

Where it shows up

Production FastAPI services with AI-integrated features at Andela.

Andela

02 / 06

AI must earn its place

Use an LLM where it creates real product value, not simply because an AI API is available.

Where it shows up

In MoodVerse, Gemini interprets the reflection and never writes the verse.

MoodVerse

03 / 06

Keep systems replaceable

Design integrations so providers, models and infrastructure can change without rewriting core application logic.

Where it shows up

Insight Stream integrates OpenAI, Gemini, DeepSeek and Groq.

Insight Stream

04 / 06

Async when it matters

Move slow or heavy work out of the request path when it improves reliability and response time.

Where it shows up

Celery and Redis run ingestion in Insight Stream and price checks in AlertChain.

Insight Stream AlertChain

05 / 06

Data before guessing

For knowledge-based AI, use retrieval and trusted data instead of asking a model to invent facts.

Where it shows up

SEO Maker’s AI explains findings from measured page data.

SEO Maker

06 / 06

Simple systems scale better

Prefer clear boundaries, observable workflows and boring infrastructure over unnecessary complexity.

Where it shows up

Workroom keeps its AI service separate from the main Django app.

Workroom

Case study 01, MoodVerse

How do you recommend scripture by mood without ever letting an LLM invent a verse?

Split interpretation from retrieval. The model reads the reflection, and a verified corpus supplies every passage.

Read the case study All case studies

More write-ups

Grounding AI recommendations in real page dataSEO Maker, published Building an AI content intelligence pipelineInsight Stream, published Building a reliable price-alert systemAlertChain, published Designing secure project-level access controlWorkroom, published

Hiring for a backend or AI role?

I’m open to remote backend, AI systems and freelance work. Tell me what you’re building and I’ll reply within 24 hours.

Let’s talk

It’s --:-- in Addis Ababaaddis0132@gmail.com

Abel Addis · Backend engineer, AI systems · Ethiopia (UTC+3)I build the systems behind AI products.I build thesystemsbehind AI products.