Wayfair logoCurrently an Extern @Wayfair
Mehdi Salhi portrait

AI & Cloud Engineer Crafting Intelligent Solutions

I'm Mehdi Salhi, an AI and Cloud Engineer with a passion for building AI-driven systems on AWS. Currently pursuing a Computer Science degree at SNHU, I integrate large language models into impactful

View my workMessage Me

Work samples

Explore my work in AI and cloud engineering, showcasing projects that leverage AWS, large language models, and cutting-edge technologies.

  • Unkommon — Serverless AI platform on AWS
    Unkommon — Serverless AI platform on AWS

    Designed and deployed a serverless AI platform on AWS running a website chatbot and a voice agent on large language models via Amazon Bedrock. Built agentic, tool-calling workflows where the assistant checks availability and books appointments through function calls, using Python and Bedrock. Engineered the system prompt, tool schemas, and guardrails (including prompt-injection defenses), and added a Trie-based intent classifier to answer common questions before the LLM, cutting latency and per-call cost. Deployed on serverless AWS (Lambda, API Gateway, DynamoDB) as infrastructure as code with AWS SAM, with a CI pipeline and automated tests.

    Amazon BedrockLLMsPythonLambdaAWS SAM
  • Company Policy RAG — Retrieval-Augmented Generation system
    Company Policy RAG — Retrieval-Augmented Generation system

    Built a RAG pipeline with hybrid retrieval (dense embeddings + BM25) and cross-encoder reranking over proprietary documents. Evaluated retrieval quality with faithfulness and context precision, both 1.00 on a 10-question test set. Containerized with Docker and deployed with a FastAPI backend on Hugging Face Spaces.

    PythonRAGDockerFastAPI

About me

I'm Mehdi Salhi, an AI and Cloud Engineer with a passion for building AI-driven systems on AWS. Currently pursuing a Computer Science degree at SNHU, I integrate large language models into impactful

I'm an AI and Cloud Engineer, as well as an AWS Certified Solutions Architect Associate, focused on creating AI-driven systems on AWS. Currently, I'm a Computer Science student at SNHU, set to graduate in August 2026 with a 3.71 GPA. I've gained practical experience in integrating large language models into production applications, building retrieval-augmented generation pipelines, and designing workflows using Amazon Bedrock on serverless AWS infrastructure. I'm proficient in Python, LLM integr

Externships

Wayfair n8n AI Agent Engineering Externship

Wayfair

Education

Southern New Hampshire University

B.S. in Computer Science · Class of 2026

Skills

Amazon BedrockLarge Language ModelsRAGHybrid SearchCross-Encoder RerankingAgentic / Tool-Calling WorkflowsPrompt EngineeringLLM EvaluationAmazon SageMakerLambdaAPI GatewayDynamoDBS3EC2IAMEventBridgeCloudWatchTerraformAWS SAMDockerGitHub ActionsCI/CDOIDCPythonBashTypeScript/JavaScriptSQLYAML
Back to work

Unkommon — Serverless AI platform on AWS

Designed and deployed a serverless AI platform on AWS running a website chatbot and a voice agent on large language models via Amazon Bedrock. Built agentic, tool-calling workflows where the assistant checks availability and books appointments through function calls, using Python and Bedrock. Engineered the system prompt, tool schemas, and guardrails (including prompt-injection defenses), and added a Trie-based intent classifier to answer common questions before the LLM, cutting latency and per-call cost. Deployed on serverless AWS (Lambda, API Gateway, DynamoDB) as infrastructure as code with AWS SAM, with a CI pipeline and automated tests.

Amazon BedrockLLMsPythonLambdaAWS SAM
View all work
Back to work

Company Policy RAG — Retrieval-Augmented Generation system

Built a RAG pipeline with hybrid retrieval (dense embeddings + BM25) and cross-encoder reranking over proprietary documents. Evaluated retrieval quality with faithfulness and context precision, both 1.00 on a 10-question test set. Containerized with Docker and deployed with a FastAPI backend on Hugging Face Spaces.

PythonRAGDockerFastAPI
View all work
Back to work

AWS Cost Watchdog — Serverless FinOps tool on AWS

Built a serverless tool that uses ML-based anomaly detection to flag unusual AWS spend, with idle-resource detection and tag governance, surfaced in a React dashboard. Provisioned with Terraform (remote S3 state) and deployed via a GitHub Actions CI/CD pipeline using OIDC; four event-driven Lambda functions on EventBridge.

TerraformLambdaEventBridgePython
View all work
Back to work
Wayfair ↗

✅ Verified by Extern · ⏱️ In progress

Wayfair n8n AI Agent Engineering Externship

Learn one of the most in-demand skills shaping the future of work: building AI agents. In this externship, you’ll step into the role of an AI Agent Developer and use n8n to design agents that think, plan, and act on real data — automating insights, decisions, and workflows. No coding experience required. By the end, you’ll create a live dashboard powered by your own AI agents, gaining hands-on experience that applies to any role or industry — from marketing and consulting to product management and data science.

AI Workflow DesignAI Agent DevelopmentPresentation SkillsWeb Scraping & Data Extraction

Overview

During my externship at Wayfair, I focused on developing AI agents using the n8n platform. This project involved designing and automating workflows to enhance data-driven decision-making. By the end, I created a live dashboard that showcased my AI agents' capabilities, equipping me with valuable skills relevant across various industries.

Wayfair n8n AI Agent Engineering Externship

What I've accomplished

I produced three polished, trend-ready moodboard prompt sets, implemented a Prompt Node in n8n, and delivered usage instructions and prompts formatted for image generators.

Project breakdown

During the module I created a first Prompt Node and authored polished prompts and three distinct trend-ready moodboard prompt sets. The submission included full prompts ready for image generators and usage instructions.

Google Docs
Access

The externship focused on shag rugs. I built n8n workflows for input routing, Amazon and social data ingestion, AI-based attribute extraction and clustering, and generated a downloadable trend report and moodboard visuals. The submission included the exported agent JSON, a PDF report, and a demo…

Google Docs
Access

During the externship I built an n8n workflow that parsed user input, fetched product data from Wayfair, Amazon, and Walmart, and merged results for AI analysis. I produced a PDF competitor analysis report for outdoor rugs.

Download
View all work