Full Stack Engineer (Automation & AI Systems)
ThoughtBins Technologies
Aug 2025 — Present
Building automation and AI-driven workflow systems for sales, HR, and CRM use-cases using n8n, LangChain, and custom backend services.
Impact
Reduced manual intervention in repetitive workflows across sales and CRM operations.
— ThoughtBins Technologies
Key Achievements
Built automation workflows handling hundreds to thousands of executions per day across multiple clients
Implemented retry and fallback mechanisms to handle unreliable third-party APIs
Integrated 8–12 external services including CRMs, email providers, and scraping pipelines
Developed AI-assisted workflows for content generation and lead qualification
Key Highlights
Hybrid architecture combining no-code (n8n) with custom backend services
Use of LangChain/LangGraph for structured AI workflows
Handling of long-running and failure-prone workflows
Multi-client workflow customization and deployment
Responsibilities
Workflow Orchestration (n8n + Custom Backend)
Designed and deployed workflow pipelines using n8n with custom Node.js services for state management, retries, and long-running task handling.
AI Agent & LLM Integration
Built LLM-powered workflows using LangChain/LangGraph for tasks like lead enrichment, content generation, and decision support.
System Integration
Integrated multiple third-party APIs (CRMs, email, scraping tools) and internal services into unified automation pipelines.
Production Debugging
Handled failures across distributed workflows, including API timeouts, inconsistent states, and partial execution issues.
Team & Organization
Team Size
6–8 engineers
Reporting To
Engineering Manager
Technologies & Tools
Technologies
Tools Used
Challenges & Solutions
Unreliable External APIs
Solution: Implemented retry strategies with backoff and partial state recovery to prevent workflow failure cascades.
State Management in n8n
Solution: Used external persistence layers and custom services to track workflow progress beyond n8n’s default capabilities.
LLM Output Reliability
Solution: Added validation layers and structured prompting to reduce hallucinations and inconsistent outputs.
Key Learnings
Practical limitations of no-code tools at scale
Designing hybrid systems combining orchestration tools and backend services
Failure handling in distributed, API-heavy workflows
Trade-offs in using LLMs in production systems