B
Middle AI Developer
Blossom
Contract type
Ongoing
Work mode
100% remote
Experience
Mid-level · 3+ years
Job description
Key details
- Design and build production LLM-powered features, including multi-step agentic workflows, RAG pipelines, and tool-augmented assistants
- Develop and maintain LangGraph agents with well-defined state management, checkpointing, and human-in-the-loop patterns
- Integrate AWS Bedrock to invoke foundation models (Claude and others), configure Guardrails, and leverage Knowledge Bases and Agents
- Architect and operate AI workloads on AWS across Lambda, ECS/Fargate, S3, DynamoDB/RDS, and CloudWatch
- Set up and manage MCP (Model Context Protocol) servers and clients to extend agent capabilities with external tools and data sources
- Own prompt engineering and evaluation: define metrics, run experiments in Braintrust, LangSmith, or Promptfoo, and iterate systematically
- Build and maintain RAG architectures: manage embeddings, vector databases (OpenSearch, pgvector, Pinecone), and chunking strategies
- Maintain high engineering standards: write typed, tested Python; enforce CI/CD pipelines; manage infrastructure with CDK, Terraform, or CloudFormation
- Work with Claude Code and comparable agentic coding tools to accelerate development — authoring CLAUDE.md files, custom slash commands, and team workflow integrations
- Monitor cost, latency, and reliability of LLM workloads; propose and implement optimizations
- Company mission
- Information not specified
Primary stack
Core technologies
.NETPython (Programming Language)AWSTypeScriptDockerNode.js
Benefits
- 100% Remote Work
- Mentorship from experienced product and banking professionals
- Multicultural collaboration with teams in Colombia and the U.S
- Real exposure to the American financial market
- Growth path into senior product or leadership roles as the product scales
Requirements & details
- 3+ years of professional software engineering experience, with at least 1 year focused on building and operating production AI/LLM applications
- Hands-on experience with Claude Code or comparable agentic coding tools: writing CLAUDE.md files, building custom slash commands, configuring MCP servers, and integrating them into team workflows
- Practical experience invoking foundation models via the Bedrock Runtime API, working with Claude models on Bedrock, and understanding Bedrock Guardrails, Knowledge Bases, or Agents
- Solid working knowledge of IAM, Lambda, ECS/Fargate, S3, DynamoDB or RDS, CloudWatch, and VPC networking. Comfortable deploying and operating services in production
- Production experience designing stateful multi-step agents: building graph workflows with nodes and edges, managing state and checkpointing, and handling tool calls and human-in-the-loop patterns
- Strong Python skills (typing, async, testing). TypeScript/Node.js is a plus
- Experience with prompt engineering and evaluation frameworks (Braintrust, LangSmith, Promptfoo), iterating on agent behavior with measurable metrics
- Understanding of RAG architectures, embeddings, vector databases (OpenSearch, pgvector, Pinecone), and chunking strategies
- Comfortable with API design, observability, and cost/latency optimization for LLM workloads
- Git, CI/CD, and Infrastructure as Code (CDK, Terraform, or CloudFormation)
- Experience building or integrating MCP (Model Context Protocol) servers and clients
- Fintech or regulated-industry experience (compliance, audit trails, data residency)
- Multi-tenant SaaS architecture design and implementation
- Experience fine-tuning or distilling language models
- LLMs, AI Agents, LangChain, OpenAI, Python (Programming Language), TypeScript, Node.js, AWS, Docker
- .NET
- Node.js
- Python (Programming Language)
- AWS
- TypeScript
- Docker