About
You'll partner with product, engineering, and operations to identify high-ROI use cases, build and deploy solutions end-to-end, and establish the foundations (architecture, evaluation, monitoring, and security) so AI is reliable at scale.
What success looks like (first 4 weeks):
1. Deliver 2–4 AI features into production that save time, reduce errors, or unlock new capabilities in Utiliko (measured with clear before/after impact).
2. Stand up a repeatable AI delivery pipeline: prompt/versioning, evaluation harness, logging/monitoring, and rollback plans.
3. Implement a secure approach to handling sensitive data (PII), with clear controls and auditability.
4. Create internal documentation so the team can maintain and extend what you build.
What you'll build
Examples of the types of projects we expect:
1. AI copilots inside ERP modules (CRM, tickets, operations, accounting support, HR support).
2. Document intelligence: extracting structured data from PDFs/emails/contracts/invoices into ERP entities with validation.
3. Workflow automation: routing, summaries, task creation, follow-ups, and exception handling.
4. Search and retrieval across internal knowledge (RAG), with strong accuracy and citations.
5. Agentic workflows (where appropriate) with guardrails: approvals, human-in-the-loop, and deterministic fallbacks.
6. Evaluation + monitoring: quality metrics, regression tests, cost/latency controls, and alerting.
Responsibilities
1. Own AI initiatives end-to-end: discovery → prototype → production → monitoring → iteration.
2. Design the AI architecture that fits a real ERP: permissions, multi-tenant patterns (if applicable), and role-based access.
3. Build reliable integrations with LLMs and internal services (APIs, event streams, background jobs).
4. Implement model/tool calling, retrieval, memory patterns (where appropriate), and structured outputs.
5. Create evaluation frameworks (offline + online), and continuously improve accuracy and reliability.
6. Ensure solutions are secure, compliant, and cost-controlled.
Required experience (non-negotiable)
To filter for real implementers, make these explicit:
1. 3+ years building and shipping software in production environments.
2. 1+ year shipping LLM-based or ML-assisted features to real users (internal or external).
3. Strong backend engineering skills: APIs, databases, background jobs/queues, caching, and observability.
4. Experience with at least one major language deeply (Python or TypeScript/Node preferred).
5. Proven ability to work from ambiguous problem → scoped plan → shipped result.
6. Comfort owning quality: tests, monitoring, incident response, and iterative improvement.
Strongly preferred:
1. Built RAG systems with real evaluation (not just demos).
2. Experience with structured extraction (OCR, document parsing), and validation workflows.
3. Experience with vector databases and hybrid search.
4. Experience with CI/CD, Docker, cloud deployments, and production monitoring.
5. Knowledge of data privacy practices (PII handling), permissions models, and security reviews.
6. Experience building internal tools / ERPs / workflow systems.
What we don't want (this role won't fit if…)
Your experience is mostly prompts in a playground with no production deployments.
You can't share examples of shipped work (screenshots, repos, architecture diagrams, metrics).
You prefer research-only work without owning deployment, reliability, and outcomes.
Interview process (proof-of-work)
We hire based on demonstrated competency.
1. Portfolio Review (Required): Walk us through 2–3 relevant projects you personally built. Be specific about your role, architecture, tradeoffs, and measurable outcomes.
2. Practical Proof of Concept (PoC): You'll implement a small scoped AI feature that mirrors our real work (structured output, permissions, evaluation plan, and basic monitoring). This is designed to show how you build, not how you talk.
3. Team + Execution Interview: Collaboration style, prioritization, and how you operate in a fast-moving environment.
Application instructions (to filter out low-signal applicants)
Include all of the following or we won't review:
1. Links to GitHub, portfolio, or deployed demos (or screenshots if private).
2. A short write-up of one AI feature you shipped: the problem, your approach, what you built, and impact.
3. Your preferred stack (Python/Node, frameworks, LLM tooling).
4. If your repos are private, be ready to screen share and walk through code.
Compensation & location
Full-time role. Compensation based on experience and demonstrated ability to ship.
Location: Remote
Contract duration of more than 6 months. with 40 hours per week.
Mandatory skills: AI Integration, LLM Prompt Engineering, LangChain, Python, API Integration
Languages
- English
Notice for Users
This job comes from a TieTalent partner platform. Click "Apply Now" to submit your application directly on their site.