27 Portfolio Projects to Move from Software Engineer to AI Engineer

Bashiri Smith · Facebook reel · 2026-09-26 · 0:08 · 18,696 views · Open on Facebook

Topics: Portfolio Projects, Resume, Job Search & Interviews · Level: intermediate

Summary

This 8-second reel has no speech. The caption lists 27 portfolio projects meant to help a software engineer move into AI engineering. They range from production infrastructure (LLM gateways, semantic caching, cost control, regression detection) to RAG, agents, fine-tuning, evals and security, plus some industry-specific ones (HVAC, insurance, legal, freight, recruiting). The creator offers a longer YouTube video covering each project's tools, how long to spend on it and how to find real users.

Key points

• The advice is to build at least one of the 27 projects to show you can do AI engineering work, not just regular software engineering.
• Production/LLMOps projects: model regression detection, LLM cost autopilot, failure forensics for AI pipelines, an LLM gateway with rate limiting and fallback routing, a self-healing LLM gateway, an AI feature flag system and a semantic caching layer.
• Retrieval projects: RAG with hybrid search, knowledge graph RAG for company data, a multimodal RAG assistant for HVAC technicians, and an evaluation system for legal research retrieval.
• Agent projects: an agent orchestration system, a multi-agent insurance claims reviewer, a multi-agent research assistant and an LLM output arbitration system.
• Eval projects: a prompt versioning + A/B testing platform, an automated eval dataset generator, and LLM-as-a-judge checked against human ratings.
• Model customization projects: a fine-tuning pipeline with LoRA, LoRA fine-tuning for freight invoice extraction, and a model distillation pipeline.
• Safety projects: Text-to-SQL with guardrails, prompt injection defense for recruiting assistants, and an automated AI red-teaming harness.
• Industry-specific projects (HVAC, insurance, e-commerce, legal, freight, recruiting) are suggested; the creator's full video says finding real users matters.

Resources mentioned

• 27 AI Engineering Projects to Get Hired - The Ultimate Guide · video · youtube.com · free
  A longer walkthrough of all 27 projects: what to build, which tools you need, how long to spend building and how to find real users.
  Also in: Software Engineer to AI Engineer Before 2027: A 5-Step Career Plan (Bashiri Smith on Facebook · notes)

Try this

• Pick one of the 27 projects and build it as a portfolio piece.
• Comment "27" on the reel to get the creator's full YouTube video on all 27 projects.
• Find real users for your project, as the full video recommends.
• Model regression detection system
• LLM cost autopilot
• Failure forensics tool for AI pipelines
• Self-healing documentation bot
• LLM output arbitration system
• RAG pipeline with hybrid search
• Semantic caching layer
• Text-to-SQL with guardrails
• Prompt versioning + A/B testing platform
• Fine-tuning pipeline with LoRA
• LLM gateway with rate limiting + fallback routing
• AI feature flag system
• Automated eval dataset generator
• Multimodal document processor
• Agent orchestration system
• Multimodal RAG assistant for HVAC technicians
• Multi-agent insurance claims reviewer
• E-commerce support caching system
• Legal research retrieval evaluation system
• LoRA fine-tuning for freight invoice extraction
• Prompt injection defense for recruiting assistants
• Knowledge graph RAG for enterprise data
• Multi-agent research assistant
• Self-healing LLM gateway
• LLM-as-a-judge with human calibration
• Model distillation pipeline
• Automated AI red-teaming harness