AI Software Engineer & Backend Developer
Engineering deterministic, production-ready AI systems, agentic workflows, and high-throughput cloud APIs.
I am an AI Software Engineer & Backend Developer with a Computer Science background (B.Sc., Mansoura University) specializing in transforming probabilistic AI models into deterministic, high-throughput production software.
Unlike experimental notebook scripts, I engineer the low-latency streaming pipelines, stateful agent graphs, and multi-tier safety guardrails needed to run AI reliably at enterprise scale:
- Agentic Multi-Step Workflows: Stateful tool-calling systems using LangGraph featuring cyclic execution graphs, memory persistence, and human-in-the-loop review nodes.
- Production Retrieval-Augmented Generation (RAG): Hybrid lexical-vector search, semantic reranking, dynamic context windows, and exact citation attribution.
- Resilient Microservices & Cloud Infrastructure: Built with Python (AsyncIO), FastAPI, Redis, and deployed with Docker, Kubernetes (EKS), and Terraform.
- AI Safety & Alignment: Practical experience evaluating frontier LLMs at Outlier/Atlas Capture, implementing multi-layer prompt injection mitigations, and enforcing strict Pydantic output validation.
| System / Project | Architecture & Engineering Highlights | Tech Stack | Live Links |
|---|---|---|---|
|
🎙️ AutoHire Autonomous AI Technical Interviewer |
• Full-duplex voice & text assessment platform with sub-300ms WebSocket streaming latency. • Multi-agent LangGraph evaluation graph scoring responses against dynamic rubrics and generating asynchronous scorecards. • Multi-tier LLM fallback routing (Groq Qwen-27b + Gemini Flash) ensuring 99.9% uptime. |
FastAPI LangGraphWebSockets RedisNext.js 16 TypeScript
|
🌐 Live Web Demo ⚡ Swagger API 📂 GitHub Repo |
|
🚢 Shiphny AI Support Logistics & Customer Operations Agent |
• Audited 5-layer prompt injection defense framework passing 24/24 penetration attack vectors. • Asynchronous task decoupling via Redis and Celery queues to maintain sub-500ms API response times under load. • Real-time parcel tracking via RAG with automated GDPR-compliant PII data masking. |
FastAPI LangChainAWS EKS TerraformRedis Docker
|
🌐 Live Demo 📂 GitHub Repo |
|
👁️ Meridian AI Auditor Multimodal Industrial Document & Audio Inspector |
• Multimodal OCR processing complex engineering schematics and invoices via Llama 3.2 Vision. • Whisper voice integration for automated audio memo transcription and compliance cross-checks. • Strict Pydantic schema validation ensuring deterministic data extraction into downstream ERPs. |
FastAPI WhisperLlama Vision PydanticPython OCR
|
🌐 Live Demo 📂 GitHub Repo |
|
🚨 Autonomous SRE Swarm Self-Healing Incident Remediation Swarm |
• Multi-agent AIOps swarm using LangGraph for log anomaly detection and root-cause analysis (RCA). • Automated canary rollback pipelines and self-healing cloud infrastructure remediation workflows. • Human-in-the-loop review gates preventing unauthorized cluster state modifications. |
Python LangGraphKubernetes AIOpsObservability Docker
|
📂 GitHub Repo |
|
🔍 ReqLens Engine Requirements Ambiguity & Clarity Analyzer |
• Automated linguistic ambiguity and conflict detection in software user stories. • Semantic requirement clustering and interactive specification generation via Gemini 2.5 Flash. • Clean Streamlit interface tailored for engineering teams and product managers. |
Python NLPStreamlit FastAPITransformers
|
🌐 Live Demo 📂 GitHub Repo |
|
🏋️ Gym & Fitness RAG Chatbot Full-Stack Intelligent Fitness Advisory |
• Hybrid lexical-vector RAG assistant providing context-grounded workout & nutritional plans. • Vector retrieval over verified fitness literature with multi-turn conversational memory in PostgreSQL. • Full-stack responsive web application built with Next.js, Tailwind CSS, and FastAPI. |
Next.js TypeScriptFastAPI RAGPostgreSQL
|
🌐 Live Demo 📂 GitHub Repo |
|
⚡ Telegram Store & Commerce Bot High-Concurrency Digital Storefront Bot |
• Asynchronous serverless bot powered by Aiogram 3, Supabase PostgreSQL, and Redis caching. • Integrated InstaPay, Vodafone Cash, and digital currency checkout flows with zero-trap FSM. • Sub-second cold starts with automated keep-alive health monitoring on Vercel. |
Aiogram 3 FastAPIPostgreSQL RedisSupabase Vercel
|
📂 GitHub Repo |
|
🛡️ AI Safety & Guardrail API High-Throughput LLM Security Firewall |
• High-speed FastAPI middleware intercepting prompts before upstream LLM inference. • Real-time prompt injection mitigation, jailbreak detection, and toxicity scoring. • Sub-millisecond latency overhead engineered for enterprise API gateways. |
FastAPI PydanticCybersecurity DockerLLM Security
|
📂 GitHub Repo |
graph LR
User[Client / WebSockets] --> Gateway[FastAPI Gateway / Rate Limiter]
Gateway --> Guardrail[AI Safety & PII Sanitizer]
Guardrail --> Router[Semantic LLM Router / LiteLLM]
Router --> Agents[LangGraph Multi-Agent Orchestrator]
Agents --> VectorDB[(Vector DB / Hybrid RAG)]
Agents --> Cache[(Redis State & Session Cache)]
Agents --> Cloud[AWS Infrastructure / Kubernetes]
- Deterministic Execution: LLMs are probabilistic; the systems wrapping them must be deterministic. Every output is strictly validated via Pydantic schemas, retry logic, and circuit breakers.
- Resilient Fallbacks: Multi-provider fallback routing (Groq, Gemini, Anthropic, OpenAI) ensures continuous uptime during provider outages or rate limits.
- Security-First Mindset: Multi-layer input sanitization, PII masking, and prompt injection filters prevent data leakage and adversarial manipulation.
| Domain | Technologies & Frameworks |
|---|---|
| AI & Agentic Systems | LangGraph • LangChain • LlamaIndex • LiteLLM • FAISS • ChromaDB • Whisper • Llama 3.2 Vision |
| Backend & Microservices | Python (AsyncIO) • FastAPI • WebSockets • Celery • Pydantic • RESTful Architecture |
| Databases & Caching | PostgreSQL • Supabase • Redis (Caching, Pub/Sub, Queues) • SQLite |
| Cloud & DevOps | Docker • Kubernetes (EKS) • Terraform (IaC) • AWS (EC2, S3, IAM) • GitHub Actions CI/CD • Linux |
| Frontend & UI | TypeScript • Next.js • React • Tailwind CSS |
|
|
|
- AI Software Engineer & Freelance Solutions Developer (Jan 2024 – Present)
- Designed and deployed production-grade RAG systems and multi-agent backend APIs using FastAPI and Python.
- Implemented multi-provider LLM fallback routing to handle upstream provider limits and maintain application reliability.
- Built automated CI/CD deployment pipelines using Docker and GitHub Actions.
- AI Alignment & Evaluation Specialist (Atlas Capture & Outlier · Dec 2024 – Jun 2025)
- Evaluated and debugged complex Python and TypeScript code generated by frontier LLMs for RLHF alignment datasets.
- Tested adversarial attack benchmarks and prompt injection mitigation to improve model safety and code accuracy.
- B.Sc. in Computer Science (Mansoura University, Egypt · 2019 – 2023)
I am actively open to AI Software Engineering roles, Backend Developer positions, and Impactful AI Projects.
- 💼 LinkedIn: linkedin.com/in/ahmed-ai-dev
- 🌐 Portfolio: ahmedgaiter.site
- 📧 Direct Email: ahmedekramy303@gmail.com
- 📍 Location: Egypt (Available for Remote worldwide & Relocation)