Asks for

The ad says“● Design, build and operate LLM- and agent-based features in production - architecture through deployment, monitoring and iteration - with cost and latency treated as design constraints, not later optimisation”
The ad says“● Integrate models with our systems and data: APIs, tool calling, multi-step orchestration, retrieval where the task requires grounding in specific data”
The ad says“● Build the evaluation and observability that make behaviour changes visible before release rather than after: test sets, automated scoring, tracing, per-task cost”
The ad says“● You have worked under data security or compliance constraints, including limits on what may be sent to third-party model providers, and you understand the failure modes specific to LLM systems - prompt injection, data leakage across tenants, unsafe tool invocation”
The ad says“● You have worked under data security or compliance constraints, including limits on what may be sent to third-party model providers, and you understand the failure modes specific to LLM systems - prompt injection, data leakage across tenants, unsafe tool invocation”
The ad says“● You have worked under data security or compliance constraints, including limits on what may be sent to third-party model providers, and you understand the failure modes specific to LLM systems - prompt injection, data leakage across tenants, unsafe tool invocation”
The ad says“● Design, build and operate LLM- and agent-based features in production - architecture through deployment, monitoring and iteration - with cost and latency treated as design constraints, not later optimisation”
The ad says“● Integrate models with our systems and data: APIs, tool calling, multi-step orchestration, retrieval where the task requires grounding in specific data”
The ad says“● Build the evaluation and observability that make behaviour changes visible before release rather than after: test sets, automated scoring, tracing, per-task cost”
The ad says“● Design for non-deterministic failure - output validation, fallbacks, degradation paths, human checkpoints where an error cannot be absorbed”
The ad says“● You have worked under data security or compliance constraints, including limits on what may be sent to third-party model providers, and you understand the failure modes specific to LLM systems - prompt injection, data leakage across tenants, unsafe tool invocation”
The ad says“● Senior-level software engineering in Python/Typescript, with production ownership: you have deployed, monitored and been accountable for systems in live use”
The ad says“● Integrate models with our systems and data: APIs, tool calling, multi-step orchestration, retrieval where the task requires grounding in specific data”
The ad says“● Build the evaluation and observability that make behaviour changes visible before release rather than after: test sets, automated scoring, tracing, per-task cost”
The ad says“● Senior-level software engineering in Python/Typescript, with production ownership: you have deployed, monitored and been accountable for systems in live use”

Tryzens Global is a digital commerce agency that helps the world's most ambitious retailers scale their businesses internationally. We are the only global support partner in digital commerce operating across every time zone, with deep experience solving complex retail integrations around the world.

We provide best of breed solutions and services to global businesses within retail, wine & spirits, beauty & cosmetics, automotive, financial services and other industries. We have built our delivery focused reputation upon technical innovation, in-depth business knowledge, and creative vision, all of which supports our objective of helping clients to gain true value from their technology stacks. We have a platform-neutral independent approach working with the world’s leading technology partners like Shopify, Agentforce(Salesforce), SAP & Adobe.

We are embedding artificial intelligence across every stage of our delivery lifecycle and are looking to grow our AI Practice in Sofia. We are looking for an experienced , smart and energetic person to join our team of talented professionals as a Senior AI Developer.

The AI Developer role is a hands-on engineering role within the Tryzens AI Practice and is focused on building AI-augmented delivery capabilities across our core technology platforms.

Responsibilities

● Design, build and operate LLM- and agent-based features in production - architecture through deployment, monitoring and iteration - with cost and latency treated as design constraints, not later optimisation

● Integrate models with our systems and data: APIs, tool calling, multi-step orchestration, retrieval where the task requires grounding in specific data

● Build the evaluation and observability that make behaviour changes visible before release rather than after: test sets, automated scoring, tracing, per-task cost

● Design for non-deterministic failure - output validation, fallbacks, degradation paths, human checkpoints where an error cannot be absorbed

● Work with clients and product owners to turn ambiguous requirements into scoped, testable capabilities, including judging when an LLM is the wrong tool

Requirements

● Senior-level software engineering in Python/Typescript, with production ownership: you have deployed, monitored and been accountable for systems in live use

● You have shipped an LLM or agent system that reached production, was used by people outside your team, and was maintained by you afterwards

● You have a working method for knowing whether a change to a prompt, model or pipeline improved the system

● You have worked under data security or compliance constraints, including limits on what may be sent to third-party model providers, and you understand the failure modes specific to LLM systems - prompt injection, data leakage across tenants, unsafe tool invocation

Useful, not required

● Retrieval at scale: hybrid search, reranking, permissioned or fast-changing document sets

● Agent frameworks, MCP

● Bedrock / Vertex / Azure OpenAI

● Asynchronous and long-running systems: queues, idempotency, durable state

● Client-facing consulting

Not required

An ML or research background. This is an engineering role: the work is building systems around models, not developing them. No specific commerce platform experience.

Benefits

● Competitive remuneration package

● Flexible working hours

● Fully remote working policy

● Food and gift vouchers

● Additional health insurance

● Great work-life balance and 25 days annual vacation

● 1 day Birthday vacation

● A positive and fun environment

Asks for

Values