class AIEngineer:
    def __init__(self):
        self.name        = "Shashibhushan Singh"
        self.role        = "AI Engineer — LLM + Backend"
        self.location    = "India 🇮🇳"
        self.experience  = ["AWS Cloud Engineer @ Amazon", "AI/LLM Engineer"]
        self.focus       = ["LLM Applications", "RAG Systems", "AI Agents", "Backend APIs"]
        self.currently   = "Building production-grade LLM backends"
        self.learning    = ["Fine-tuning (LoRA/QLoRA)", "Multi-agent systems", "MLOps"]
        self.ask_me      = ["LangChain", "FastAPI", "Vector DBs", "AWS", "Prompt Engineering"]
        self.contact     = "itshashi.io@gmail.com"

    def __str__(self):
        return "From cloud infrastructure to intelligent AI systems 🚀"

Designed, deployed, and maintained scalable cloud infrastructure on AWS

• Architected multi-region, highly available systems on EC2, ECS, Lambda, and EKS
• Built automated CI/CD pipelines using CodePipeline, CodeBuild, and GitHub Actions
• Managed data infrastructure with S3, RDS, DynamoDB, and Redshift
• Implemented infrastructure-as-code with Terraform and AWS CloudFormation
• Set up observability stacks using CloudWatch, X-Ray, and Grafana
• Enforced security best practices: IAM roles, VPC design, KMS, GuardDuty
• Reduced infrastructure costs by optimizing Reserved Instances and Auto Scaling policies

Building production AI systems powered by large language models

• Designing RAG pipelines over enterprise document corpora with retrieval + reranking
• Building LLM agents with tool use, persistent memory, and multi-step reasoning
• Shipping async FastAPI backends with streaming (SSE/WebSocket) for real-time AI
• Integrating OpenAI, Anthropic Claude, and open-source models (Mistral, LLaMA)
• Implementing LLM observability with LangSmith and Langfuse
• Exploring LoRA/QLoRA fine-tuning for domain-specific model adaptation

📄 RAG Pipelines         →  Ingest docs · Chunk · Embed · Retrieve · Rerank · Answer
🤖 AI Agents             →  Tool use · ReAct loops · Multi-agent orchestration
⚡ LLM APIs              →  FastAPI · Streaming (SSE) · Async · Auth · Rate limiting
☁️  Cloud Infrastructure  →  AWS multi-region · IaC · Auto-scaling · Cost optimization
🧪 Eval Systems          →  RAGAS · LLM-as-judge · Regression testing
🔐 Safe AI               →  Guardrails · Prompt injection defense · Output validation
📦 Fine-tuning           →  LoRA · QLoRA · PEFT · Instruction datasets

End-to-end RAG pipeline over enterprise documents

• PDF/HTML ingestion → semantic chunking → Qdrant vector store
• Hybrid search (BM25 + dense) + Cohere reranking
• Streaming FastAPI backend with LangSmith tracing
• Deployed on AWS ECS + S3 with CloudWatch monitoring
• FastAPI LangChain Qdrant OpenAI Docker AWS

Multi-tool LLM agent with persistent memory

• ReAct loop with web search, code execution, and DB query tools
• Long-term memory via Redis + embeddings
• Streamed responses over WebSocket
• Serverless deployment on AWS Lambda + API Gateway
• LangGraph FastAPI Redis OpenAI PostgreSQL AWS

Unified API gateway across OpenAI, Anthropic, and open-source models

• Smart routing by cost, latency, and capability
• Semantic caching to cut costs by ~40%
• Per-user rate limiting and API key management
• Infra managed via Terraform on AWS ECS + ElastiCache
• LiteLLM FastAPI Redis Docker Terraform AWS

AWS cost analysis and auto-optimization tool (from Amazon experience)

• Automated detection of idle EC2, oversized RDS, and unused Elastic IPs
• Scheduled Lambda jobs to snapshot and right-size instances
• Savings report dashboard via CloudWatch + QuickSight
• Python AWS Lambda Boto3 CloudFormation CloudWatch

✅  AWS Cloud Engineer @ Amazon
     └─ EC2 · ECS · Lambda · EKS · S3 · RDS · DynamoDB
     └─ Terraform · CloudFormation · IAM · VPC · CloudWatch

✅  LLM Foundations
     └─ Transformers · Attention · Tokenization · Embeddings

✅  LLM APIs & Prompt Engineering
     └─ OpenAI · Anthropic · Few-shot · CoT · Structured outputs

✅  RAG Systems
     └─ Chunking · Vector DBs · LangChain · LlamaIndex · Reranking

✅  FastAPI AI Backends
     └─ Async · Streaming · WebSockets · Pydantic · Auth

🔄  Fine-tuning (LoRA / QLoRA)         ← currently here
🔄  Multi-agent systems (LangGraph, CrewAI)

⬜  LLM inference optimization (vLLM, TensorRT-LLM)
⬜  MLOps at scale (BentoML, Kubeflow)
⬜  Multimodal models (vision + language)

I'm open to collaborating on LLM projects, AI tooling, cloud architecture, and backend systems.
Building something interesting? Let's talk.

"Cloud gives you the scale. AI gives you the intelligence. Together — that's the future."