01 · Distributed work
Slow work leaves the request; failures are retried and recovered.
queue
- worker 1done
- worker 2running
- worker 3retry 2/3
APIqueueworkersretry
One UPI search checks a plot against 11 national map layers, and says “Unknown” when a source has no data.
Next.js · PostGIS · ArcGIS REST
Read case studyUploads return straight away; Celery workers retry a failed job 3 times with backoff before marking it FAILED.
FastAPI · Celery + Redis · PostgreSQL
Read case studyCut failed jobs 60% with retry and recovery in the document pipeline, with APIs under 200 ms.
React + TypeScript · FastAPI · PostgreSQL · Azure Service Bus · LLM APIs
Read case study−60%failed jobs after retry and recovery work
One UPI search checks a plot against 11 national map layers, and says “Unknown” when a source has no data.
Next.js · PostGIS · ArcGIS REST
Read case studyUploads return straight away; Celery workers retry a failed job 3 times with backoff before marking it FAILED.
FastAPI · Celery + Redis · PostgreSQL
Read case studyCut failed jobs 60% with retry and recovery in the document pipeline, with APIs under 200 ms.
React + TypeScript · FastAPI · PostgreSQL · Azure Service Bus · LLM APIs
Read case studyUnknown
shown when a source has no data, never a guess
11
national map layers checked in one UPI search
4 pages
PDF report, with the source and date of each fact
3
retries with backoff before a failed job is marked FAILED
300
pages a document can have and still be questioned, using retrieval
6
services started with one docker compose up
−60%
failed jobs after retry and recovery work
<200 ms
API response while document analysis ran in the background
14
document and language features on one shared platform
Hope TuyishimeHello, I'm Hope
Most recently at IFAD, a UN agency in Rome, I built asynchronous document-processing services with FastAPI, PostgreSQL and Azure Service Bus. Now I'm building Trustplot. Next, I want a backend team where reliability and data are the job.
Now
Since May 2026Land and property intelligence for Rwanda. One search checks a plot against 11 national map layers.
Read the case studyBased in
Central Africa Time · UTC+2
Open to backend engineering roles
What I handle well
01 · Distributed work
queue
APIqueueworkersretry
02 · Backend
POST /documents
201 Created
{ "status": "QUEUED" }
03 · Data
04 · AI
05 · Shipping
Designed and built a land and property intelligence platform for Rwanda: UPI parcel search, interactive mapping, zoning and risk screening, reference valuation and automated reports, on PostGIS and ArcGIS data.Case study
Graduated in software engineering with Distinction.
Built FastAPI and PostgreSQL services for document ingestion, OCR, extraction and review (~100 documents a day), with Azure Service Bus workers that kept APIs under 200 ms and cut failed jobs by 60%.Case study
Developed Java, Spring Boot and PostgreSQL REST APIs for academic and financial management systems, with relational models and tuned SQL for transactional workloads.
Built Node.js and TypeScript services and event-driven pipelines for real-time telemetry, with secure REST APIs and access control between distributed components.
Developed multi-tenant REST APIs for authentication and authorisation, improved database performance with indexing and SQL optimisation, and contributed GitHub Actions CI/CD.