AI Solution Architect

Who are We
mindit.io is an AI-driven custom software services company with +250 IT specialists with proven expertise in software & product engineering, artificial intelligence, business intelligence, data analytics & integration services.
With offices in Switzerland, Germany and 4 delivery centers in Romania, we cover the entire development cycle from business research, consulting, and architecture to post-implementation testing and maintenance.
Our Values
Partner of choice: With a customer-centric approach, we build an environment where partnerships thrive and turn them into long-lasting relations. It’s a space of co-creation, innovation, and top-notch technologies to deliver outstanding custom software development
Professional: We always deliver and this is one non-negotiable pillar of our culture. Be it product engineering or custom software development, the quality and ethics of our work are inherent to everything we do
Authentic: We value everyone’s uniqueness in the pursuit of framing a culture of equality, diversity, and inclusion. Come as you are and stay true to yourself. We become the best version of ourselves together
Team: Trust is our ultimate currency, and we use it to constantly enrich our kind community. When it comes to scaling, the team is our functional cell. We grow together, we learn together, we build together
Role Summary
The AI Solution Architect leads presales solutioning and technical delivery for AI-driven engagements by rapidly designing high-quality, client-focused AI solutions. The role merges classic solution architecture responsibilities (presales, high-level design, technical leadership, escalation, mentorship) with deep, hands-on expertise in generative AI, agentic systems, and governable AI.
At mindit.io, architects are hands-on. The AI Solution Architect must be able to write code, perform low-level design, prototype agents, pipelines, and validate low-level designs architected by software and AI engineers. They are equally credible whiteboarding an enterprise AI governance model for a CIO and debugging an agent orchestration loop with an engineer.
The defining mindset of this role is the ability to design for uncertainty. Traditional software executes deterministic paths; agentic systems operate on goals, and the exact path they take may change with every execution. The AI Solution Architect designs outcomes, boundaries, tools, and success criteria (not step-by-step workflows) and architects systems that gracefully absorb hallucinations, loop failures, and unexpected agent behavior without breaking downstream business processes.
A second defining trait is a thorough understanding of the governable AI landscape, responsible AI frameworks, model risk management, data governance, auditability, observability, and safety mechanisms (and the ability to translate these into concrete architecture decisions that enterprise clients (including regulated industries such as banking) can trust and adopt).
Key Responsibilities
Presales: fast, high-quality AI solutioning and offering
Leads presales involvement for AI opportunities: creates solutions and offers, answers presales questions with authority on AI architecture, feasibility, cost, and risk.
Reviews Business Requirements Documents (BRD) and reframes business problems as AI-solvable (or explicitly not AI-solvable) use cases.
Delivers High-Level Solution Architecture (HLSA) for AI solutions during presales, covering model selection, agent/RAG design, data architecture, integration, security, governance, and observability.
Proposes infrastructure dimensioning and cost modeling for AI workloads (compute/capacity, model consumption/token economics, storage, etc), explicitly modeling the "thinking budget" of agentic workflows (including the honest recommendation when a single-shot prompt or plain code beats a multi-agent loop on ROI).
Offers support and technical leadership in answering RFPs, with a specific ability to address AI governance, compliance, data residency, and responsible AI questions credibly.
Leads and implements small POCs and MVPs (hands-on) to de-risk proposals and demonstrate value quickly.
Advises clients on build-vs-buy decisions across the AI platform landscape.
Delivery: technical excellence
Is knowledgeable of all AI projects in their area of responsibility and their Technical Leads (TLs).
Is the escalation point from TLs for AI architecture, quality, performance, and safety concerns.
Owns all AI projects in the area: awareness, stakeholder knowledge, support for Team Managers and Portfolio Managers.
Validates low-level designs produced by engineers for AI components: agent workflows, state management, orchestration logic, retrieval pipelines, prompt and context architecture, evaluation harnesses.
Sets and enforces production standards for AI systems: latency, cost efficiency, reliability, hallucination mitigation, guardrails, monitoring, and documentation.
Helps Delivery Managers and Account Executives identify and shape upsell opportunities (e.g., extending initial pilot concepts into an agentic workflow, adding governance layers, scaling a POC to production).
Helps Delivery Managers and Team Managers with cross-sell into adjacent AI use cases across client organizations.
Can play the role of Technical Lead on specific strategic AI projects.
AI practice: thought leadership
Monitors the fast-moving AI landscape (model releases, agent frameworks, platform capabilities) and translates it into actionable guidance for teams and clients.
Forms and engages with engineering communities internally and externally; represents mindit.io's AI expertise with clients and in public forums.
Provides technical leadership to Technical Leads, including monthly calls; nurtures TLs' AI expertise and involves them in presales activities.
Is an internal thought leader, strengthening mindit.io's technical DNA in AI.
Is the main promoter of AI engineering best practices across the organization.
Sets milestones and helps Technical Leaders own and lead their projects.
Sets and follows technical quality standards for AI projects in their area.
Organizes and leads technical training initiatives (AI learning paths, workshops, enablement for engineering and business teams).
Mentors senior and junior engineers transitioning into AI engineering roles.
Skills And Qualifications
Required Qualifications
Architecture & engineering foundation
Track record: 8+ years in software engineering, of which 3+ years in an architecture or technical leadership role.
Hands-on coding: proven ability to write production-quality code in at least one of Python, TypeScript/Node.js, or Java.
Low-level design: ability to produce low-level designs and to validate those produced by engineers.
Integration architecture: robust knowledge of REST, GraphQL, event-driven architecture, microservices, and enterprise integration patterns.
Cloud foundation: strong grounding in cloud architecture, security, identity, and networking.
Agentic mindset — designing for uncertainty
Non-deterministic execution: ability to design systems judged on the outcomes they guarantee, not the steps they prescribe — an agent pursues a goal, and the same input may take a different path on every run.
Outcomes over workflows: ability to specify the goal, the boundaries, the tools the agent may call, and the success criteria (leaving the path between them to the agent).
Failure as a design input: experience building fallbacks, retries, evaluation gates, and safe degradation paths that absorb hallucinations, loop failures, and unexpected agent behavior without breaking downstream business processes.
Autonomy calibrated to risk: judgment to place each use case on the ladder from suggestion, to approval-gated action, to unattended execution — and to recognize when deterministic code is the better answer.
Generative & agentic AI
LLM systems delivery: has taken at least 2–3 LLM-powered systems (copilots, assistants, semantic search, RAG pipelines, or autonomous/semi-autonomous agents), from design to production.
Agentic patterns: command of the following architecture patterns and the judgment to know when each fits the problem's complexity (and when it is overkill):
Reflection & self-correction: loops where an agent evaluates its own output against a rubric before passing it downstream.
Plan-and-solve: two-tiered systems where a planner agent decomposes a complex goal into tasks executed by specialized execution agents.
Multi-agent orchestration: specialized, collaborating agents (e.g., researcher, coder, reviewer) with clear state hand-off contracts.
Human-in-the-loop (HITL): human approval gates and escalation paths for high-stakes decisions (financial transactions, data deletion, external communications).
AI interoperability & memory
Tool use & API integration: designs robust, securely scoped APIs specifically for AI consumption; fluent in interoperability protocols such as the Model Context Protocol (MCP) and Agent-to-Agent (A2A) standards.
Context & memory management: architects stateful agent memory beyond traditional databases — vector databases for semantic search and RAG, episodic memory (what the agent di

Work arrangement
Yes