Kozak Agency

We are looking for a skilled fullstack developer to support the further development of a production-grade web application. The stack includes Angular (latest version) as the frontend framework, NestJS for the backend, MariaDB as the database, and Keycloak for authentication and authorization. We use TypeORM for data access.

Responsibilities:

• Build and maintain Angular components (standalone components, Signals)
• Extend and improve REST APIs using NestJS
• Manage the MariaDB database schema and migrations using TypeORM
• Implement and maintain Keycloak-based authentication, including role-based access control
• Write end-to-end tests using Playwright or Cypress
• Ensure high code quality and contribute to technical documentation

Requirements:

• Strong experience with Angular (v16 or newer), including Signals
• Proficiency with NestJS and modular backend architecture
• Solid TypeScript knowledge
• Hands-on experience with MariaDB and TypeORM
• Real-world experience integrating Keycloak into fullstack apps
• Experience with Playwright or Cypress for end-to-end testing
• Familiarity with Docker, Git, and CI/CD pipelines

Project Info:

• The application is already live and actively used
• Tasks are well-defined and managed via GitHub
• Collaboration with an internal development team
• Modern, containerized infrastructure
• We’re looking for someone who can get started soon. If the collaboration works well, long-term cooperation is possible.

We’re looking for an Applied AI Engineer who combines strong ML fundamentals with the discipline of improving production AI systems through metrics, evaluation, and iteration.

Details:
Senior AI/ML Engineer, ideally 5 years of experience. Ready to consider candidates starting with 3 years of experience, but they must be very strong in AI/ML.
Strong proficiency in English (Upper-intermediate level or Advanced)
Full-time
Location: any
Start: ASAP!

The role in a nutshell
You’ll work on improving production AI systems through evaluation, experimentation, and system design.

A large part of the role involves:

diagnosing failures in agent workflows
designing evaluation metrics and KPIs
improving system prompts and agent behavior
running structured experiments and measuring impact
You won’t be working in isolation on research projects — you’ll be improving systems that real users depend on.

Rough responsibility breakdown:

AI evaluation and KPI design — ~30%
Prompt and agent system design — ~30%
ML systems (recommendation, optimization, etc.) — ~30%
Engineering integration — ~10%

What you’ll work on:

AI evaluation and system quality

Design evaluation strategies for LLM and agent workflows
Create metrics and KPIs for AI system performance
Build and maintain evaluation datasets
Debug production AI failures systematically
Compare system behavior against baselines
This is a core responsibility of the role.

Multi-agent AI systems

Improve agent orchestration and workflows
Diagnose failures across agent pipelines
Refine system prompts and agent interactions
Improve reliability, latency, and response quality of ML and AI systems

You’ll contribute to areas such as:

Recommendation systems (ranking and personalization)
Itinerary optimization and constraint-based planning
LLM-based reasoning systems
Optional: computer vision pipelines
Depth in one of these areas is more important than superficial experience in all of them.

Engineering collaboration
We use:

Golang (primary production language)
Python when necessary for ML workflows
Postgres, Redis, and internal services
You don’t need to be a Go expert on day one, but you should be comfortable reading and modifying production code.
Backend engineers handle infrastructure-heavy service development — your focus is on AI system behavior and correctness.

What we’re looking for
Must-haves
1. Strong AI/ML fundamentals
You understand the theory behind what you build and can choose appropriate methods for a problem.
Examples:

evaluation metrics (precision/recall/F1/etc.)
ranking and recommendation concepts
embeddings and similarity
experimentation methodology

Not required:
academic publications
advanced theoretical math
large-scale model training experience

2. Evaluation-driven mindset You:

think in metrics and baselines
design experiments instead of guessing
measure system improvements quantitatively
debug failures methodically
This is the most important signal for the role.

3. Experience with LLM systems you’ve worked with:

prompt design
agent workflows
evaluation of LLM outputs
production LLM integrations

4. Ability to ship production systems
You can:

turn ideas into working systems
iterate based on results
balance exploration with delivery

5. Programming ability You’re comfortable writing production code in at least one language (Python, Go, or similar) and learning others when needed.

Strong signals (nice to have)
Experience improving an AI system after deployment
Recommendation systems or ranking experience
Optimization or constraint-based systems
Computer vision experience
Experience building evaluation frameworks
Golang experience
Startup or small-team engineering experience

Responsibilities

• Deployment and maintenance of production Kubernetes clusters (Kubespray)
• Ensuring High Availability (HA) for control plane, etcd, and stateful services
• IAM and security management (Keycloak, RBAC, Network Policies)
• Setup and maintenance of monitoring, logging, and alerting systems
• Support of development environments (Minikube)
• Incident response and participation in platform stability improvements
   

Requirements

Must have:

• Strong hands-on experience with Kubernetes in production environments
• Practical experience with Kubespray
• Docker / containerd
• Experience with stateful workloads (PostgreSQL, Redis)
• Helm, Git, Bash
• Monitoring stack experience (Prometheus, Grafana)
• Solid understanding of Kubernetes networking (Ingress, Services, NetworkPolicy)
   

Nice to have:

• Experience with Keycloak (OIDC, LDAP, RBAC)
• Experience with Longhorn or Rook-Ceph
• Experience with EFK stack (Fluentd, Elasticsearch, Kibana)
• Experience with HashiCorp Vault
• Experience with MinIO / S3-compatible storage
• Understanding of SRE practices (SLA / SLO principles)