Agentic AI & Automation Engineer
I design retrieval‑augmented, agentic LLM workflows that turn manual, repetitive processes into observable, reliable automation pipelines.
- Agentic AI orchestration (multi-step tools, guardrails, memory bounds)
- Retrieval-Augmented Generation (chunking, embedding relevance, fallback)
- Workflow automation (n8n / Make) bridging APIs, webhooks, data stores
- Cost & latency optimization (prompt shaping, caching, batching)
- Observability (token + latency logging, error taxonomy, evaluation loops)
LLM / Agentic: LangChain · RAG · OpenAI GPT · Embeddings · Whisper
Automation: n8n · Make.com · Webhooks · Scheduling · Error Handling
Build / Data: Python · SQL · REST APIs · JSON · Data Parsing
Storage: MySQL · PostgreSQL · MS SQL Server · Baserow · Vector DB (Chroma/Pinecone basics)
Interfaces / Tools: Streamlit · Slack · Notion · Git · GitHub Copilot
Practices: Prompt Design · Tool Chaining · Retrieval Evaluation · Cost Optimization
| Project | Description | Stack | Impact |
|---|---|---|---|
| Agentic Voice Ordering Assistant | Multi-turn voice agent w/ RAG + STT/TTS | LangChain · Whisper · OpenAI | Faster, consistent order capture |
| Workflow Automation Hub | Cross-tool hiring/ops orchestration | n8n · Baserow · Notion | Reduced manual triage time |
| News Classifier App | Lightweight BERT+LSTM inference UI | Python · Streamlit | Production-ready demo |
| Retrieval Eval Lab | Scripts to benchmark strategies | Python · LangChain | Data-driven retrieval tuning |
| Token & Cost Dashboard | Usage + latency tracking panel | Streamlit · API Logs | Visibility into spend |
Focused on: Retrieval, Agents, Orchestration, Observability
Context → Retrieve → Reason → Act → Evaluate → Iterate
- Guardrails + schema validation before tool execution
- Retrieval quality monitored (precision / recall samples)
- Memory windows trimmed; sensitive data minimized
- Retry + backoff for transient external errors
- Token + latency metrics logged for trend analysis
⭐ If something here helps you, feel free to star a repo or open a discussion.