I design and build production-style AI systems and internal tools, focusing on FastAPI backends, LLM/RAG pipelines, multimodal AI services, and cloud-native deployment.
I bring 10+ years of professional software and systems engineering experience from large-scale, safety-critical environments (Valeo, Garrett), and over the past year I’ve been focused full-time on applied AI and cloud-based ML engineering.
My work emphasizes clean architecture, reliability, explainability, and real-world constraints — not demo-only AI.
- Backend-first platforms using FastAPI, PostgreSQL, and async pipelines
- API-driven web UIs (React / Next.js) for internal tools and analytics
- Clear separation between data, AI reasoning, and application logic
- Retrieval-Augmented Generation (RAG) pipelines
- Explainable ranking and recommendation systems
- LLMs used as augmentation layers, not single points of failure
- Vision–Language Model (VLM) based image reasoning
- Document analysis with structured outputs and grounded explanations
- Robust handling of uncertainty and failure modes
- Dockerized services and CI/CD pipelines
- Azure ML pipelines and serverless APIs
- Production-style deployment and observability patterns
Backend-first AI platform that ingests, normalizes, and analyzes job postings to produce explainable recommendations and market intelligence.
- FastAPI backend with async ingestion and analysis pipelines
- PostgreSQL persistence with normalized schemas and historical snapshots
- LLM-assisted enrichment and RAG-style reasoning
- Thin React / Next.js UI consuming backend APIs
- Designed for reliability, idempotency, and explainability
https://github.com/msaleh1888/job-market-intelligence-platform
Production-style multimodal AI service for image and document analysis with grounded explanations.
- FastAPI service using Vision–Language Models (VLMs)
- Image + prompt → multimodal reasoning → structured results
- Clear separation between perception, reasoning, and interpretation
- Confidence signals, explanations, and recommended next steps
https://github.com/msaleh1888/multimodal-visual-inspection-api
Invoice → structured JSON using Azure Functions and Azure Document Intelligence, with CI/CD and monitoring.
https://github.com/msaleh1888/azure-serverless-invoice-extraction
End-to-end ML pipeline using Azure ML Batch Endpoints and reproducible YAML deployments.
https://github.com/msaleh1888/azure-ml-customer-segmentation
Document ingestion, embeddings, vector search, and grounded /ask endpoint.
https://github.com/msaleh1888/rag-llm-fastapi-microservice
Languages
Python, SQL
Backend & APIs
FastAPI, REST, async programming
LLMs & AI
RAG pipelines, prompt engineering, multimodal AI (VLMs), embeddings
Data & Storage
PostgreSQL, SQLAlchemy, Alembic, Pandas
Cloud & DevOps
Docker, CI/CD (GitHub Actions), Azure ML, Azure Functions
ML & CV
PyTorch, torchvision, scikit-learn, transfer learning
- Full-Stack AI Software Engineering roles
- Internal AI platforms and AI Factory teams
- LLM-powered tools for analytics, automation, and decision support
- Applied AI roles focused on shipping real systems, not research demos
- Email: mahmoud.saleh1888@gmail.com
- LinkedIn: https://www.linkedin.com/in/mahmoud-saleh-ba57886a/
- Portfolio: https://msaleh1888.github.io/portfolio-site/
If you’re building AI-powered software products or internal platforms, I’d love to connect.