GenAI Solution Architect

This is a remote position.
We are seeking a very hands-on GenAI Architect to take sole architectural ownership of a live, commercially deployed agentic product. A global consumer brand has already launched a GenAI-powered front-end agent in the UK, enabling customers to discover and explore products through natural language conversation. The platform is now entering a rapid multi-market roll-out: Germany first, then the Czech Republic, then Romania, with Poland and other markets running in parallel.This is a backfill in a formed, functioning squad - the product is in production and the delivery rhythm is established. You are joining to strengthen the team, not to rebuild it.
About the role:You will be embedded in the development team, and your job is to stay ahead of it. New feature requirements arrive incomplete: you are the person who discovers the unknowns inside them, identifies where the technology risk actually sits, and resolves both before they turn into development issues mid-sprint.You prove your designs rather than presenting them. Architectural decisions here are validated by delivering a working proof of concept, which the development team then takes forward to full implementation. The measure of success is simple: developers move faster because you were there first.
Tech stack:
Python, LangChain for LLM orchestration, LangGraph, LangSmith

AWS Bedrock for model access (Anthropic among the providers), with Bedrock Knowledge Bases and Bedrock Guardrails

AWS: ECS, Lambda, S3, API Gateway, DynamoDB, Cognito, EventBridge, WAF

Twilio (telephony, in production)

MCP and Agent-to-Agent integrations

What you'll be doing:
Owning the AI architecture underpinning the multi-market roll-out;

Discovering the unknowns in new feature requirements — surfacing ambiguity, hidden assumptions, and unstated constraints before implementation starts;

Identifying and mitigating technology risks early, well before they surface as development problems;

Proving designs by building proofs of concept, and handing them to the development team as the basis for full implementation;

Shaping the evaluation and observability architecture for the agentic system — this is a new area of the solution, and you will be expected to bring options and a point of view rather than inherit a design;

Designing and iterating on multi-agent architectures, and defining integration standards for MCP servers and agent-to-agent communication;

Working with the senior technical lead in Poland and the development team in India, supporting implementation day to day;

Collaborating with the data and domain architects on the wider programme to deliver joined-up solution designs;

Working closely with the Business Analyst to validate business designs and provide technical feedback;

Advising on prompt engineering strategy, tool use design, and agent reliability across multiple languages and markets.

Requirements

Advanced, production-level experience with LangChain, LangGraph, and LangSmith in Python;

Hands-on experience implementing Model Context Protocol (MCP) and Agent-to-Agent (A2A) integrations;

Solid experience with AWS, ideally including Bedrock, and comfort designing within a serverless and container-based AWS estate;

A genuinely hands-on approach — you build the proof of concept yourself rather than delegating it;

A considered view on GenAI evaluation and observability, and the ability to argue for an architectural approach in front of technical stakeholders;

Proven ability to work with under-specified requirements: extracting the real need from stakeholders and making defensible decisions under uncertainty;

Sound judgement in technology selection, with a track record of de-risking choices early rather than discovering problems in development;

Ability to work effectively with a distributed, cross-timezone team and to communicate designs clearly to mid-level and junior engineers;

Strong communication skills — able to translate complex AI concepts for technical and non-technical stakeholders;

English proficiency at B2 level or above.

Nice to have:
Experience with telephony and call centre applications, particularly Twilio and voice/AI integration patterns;

Experience scaling an AI product across multiple languages or markets;

Background in enterprise AI platform architecture.

Originally posted on Himalayas