Looking for a job

Yosri Khiari

@yosrikhiari

Message

Software engineer at Neoledge MEA who built an OCR-to-RAG document pipeline and an 867-case quality gate for AI search.

Tunisia

Message

What I'm looking for

I'm keen to work where data quality decides model quality.

At Neoledge MEA, I built the ingestion pipeline, retrieval backend, and automated quality gate for a semantic document search engine. Its 867-case evaluation harness checks retrieval quality, answer grounding, and multilingual query parsing on every commit, blocking merges when quality falls below thresholds.

At HorizonLux, I owned rule-heavy backend modules for a multi-tenant ERP, including Tunisian tax declarations and tenant-isolation checks in CI. At L-Mobile, I turned a Prophet and XGBoost forecasting model into a pipeline planners could retrain from the UI, and built a warehouse-routing application that improved logistics efficiency by 25%.

Experience

Work history, roles, and key accomplishments

Software Engineer Intern

Neoledge MEA

Feb 2026 - Aug 2026 (6 months)

Built the ingestion pipeline, retrieval backend and automated quality gate of a semantic document search engine. Developed an 867-case evaluation harness that blocks merges below quality thresholds and implemented async OCR/NER pipelines with RabbitMQ and OpenSearch.

Python ASP.NET Core OpenSearch RabbitMQ Apache Tika Tesseract GLiNER RAG Graph RAG PostgreSQL Docker Kubernetes OpenTelemetry Prometheus Grafana

Full-Stack Developer Intern

HorizonLux

Sep 2025 - Jan 2026 (4 months)

Owned rule-heavy backend modules of a multi-tenant ERP, including a compliance engine for Tunisian tax declarations and expense management workflows. Implemented tenant isolation with PostgreSQL Row-Level Security and built a self-verifying LLM pipeline with fallback across four providers.

.NET React TypeScript PostgreSQL Row level security EF Core OCR LLM CI CD GitHub Actions

ML Engineer Intern

L-Mobile

Jul 2025 - Aug 2025 (1 month)

Turned a Prophet and XGBoost demand-forecasting model into a retrainable pipeline for planners. Orchestrated feature extraction and training from ASP.NET Core with Redis caching and monitoring via Prometheus and Grafana.

Python ASP.NET Core Redis Prometheus Grafana Azure Container Registry Prophet XGBoost