Senior Artificial Intelligence/Machine Learning Engineer

Ciklum is looking for a Senior Artificial Intelligence/Machine Learning Engineer to join our team full-time in Ukraine.

We are a custom product engineering company that supports both multinational organizations and scaling startups to solve their most complex business challenges. With a global team of over 4,000 highly skilled developers, consultants, analysts and product owners, we engineer technology that redefines industries and shapes the way people live.

About the role:

As a Senior Artificial Intelligence/Machine Learning Engineer, you will lead the technical implementation of an agentic document extraction system on AWS, anchored in Amazon Bedrock. The system ingests documents of any image quality, including scans-of-scans, and produces a structured abstract with per-field confidence scores and an automated flagged-for-review queue.  You will work closely with AI Architects & Data Engineers to design ingestion, extraction, confidence-scoring, and escalation patterns that future document agents can reuse.

Project Description:

Ciklum is seeking a Senior AI Engineer to build a production-grade document intelligence platform: an agentic system that ingests PDFs of arbitrary quality and produces structured records matching a target schema, with per-field confidence scoring and human-in-the-loop review. This role is central to establishing a production-ready foundation that handles the full pipeline from ingestion through extraction, confidence routing, and the feedback loop. You will be responsible for building high-performance extraction agents and the underlying architecture that powers them.

Responsibilities:

• Agent Development: Design, build, and test production-grade extraction agents that map unstructured document language to a structured target schema, supporting multiple schema variants
• Ingestion Engineering: Build a document ingestion pipeline robust to degraded inputs, including low-quality scans, multi-generation photocopies, mixed image and text PDFs, and inconsistent layouts
• Confidence & Routing: Develop a per-field confidence model that drives routing, where high-confidence fields flow straight through and low-confidence fields land in the review queue
• Knowledge Engineering: Build and configure RAG knowledge bases to provide agents with secure, domain-specific context, controlling grounding and hallucination on documents where a wrong value has real downstream consequences
• Safety & Governance: Implement Human-in-the-Loop escalation interfaces, confidence thresholds, and egress controls, where reviewer corrections feed back as training signal
• Re-processing: Build a diff path that compares new output against prior records and surfaces material changes
• Optimization: Configure model selection, token optimization, and cost and latency strategies across the inference pipeline
• Mentorship: Provide technical guidance to mid-level AI engineers and standardize extraction, confidence, and orchestration patterns

Requirements:

• AWS & Bedrock: Deep AWS architecture experience with hands-on Bedrock work, including model selection, prompt and context engineering, evaluation, and cost and latency optimization in production
• Document AI: Strong background in OCR on degraded inputs (Textract or equivalent), layout-aware extraction, and handling the gap between what OCR returns and what a schema needs
• Structured Extraction: Demonstrated experience extracting structured data from long, unstructured documents into a fixed schema, including confidence estimation and calibration
• Agentic Intelligence: Proven track record building multi-step reasoning chains, tool-use protocols, and state management for autonomous agents
• Pipeline Engineering: Production RAG or LLM pipeline experience, including chunking, retrieval, and grounding
• HITL Systems: Ability to design human-in-the-loop systems where model output is triaged, not blindly trusted, and reviewer corrections become training signal
• Ownership: Comfort owning a system end to end, from ingestion and inference to evaluation harness, monitoring, and the feedback loop

Desirable:

• Confidence Calibration: Selective-prediction techniques and knowing when the model doesn't know
• Evals Literacy: Experience building programmatic evaluation harnesses and testing against golden datasets to measure field-level accuracy, review-queue precision and recall, and drift
• Data Governance: Experience building on proprietary domain corpora

What’s in it for you?

• Strong community: Work alongside top professionals in a friendly, open-door environment
• Growth focus: Take on large-scale projects with a global impact and expand your expertise
• Tailored learning: Boost your skills with internal events (meetups, conferences, workshops), Udemy access, language courses, and company-paid certifications
• Endless opportunities: Explore diverse domains through internal mobility, finding the best fit to gain hands-on experience with cutting-edge technologies
• Flexibility: Enjoy radical flexibility – work remotely or from an office, your choice
• Care: We’ve got you covered with company-paid medical insurance, mental health support, and financial & legal consultations

About us:

At Ciklum, we are always exploring innovations, empowering each other to achieve more, and engineering solutions that matter. With us, you’ll work with cutting-edge technologies, contribute to impactful projects, and be part of a One Team culture that values collaboration and progress.

As one of Ukraine’s largest IT companies and a top employer recognized by Forbes, we’ve spent over 20 years delivering meaningful tech solutions. We proudly support diverse talent and military veterans, recognizing their unique skills and perspectives they bring to shaping the future.

Explore, empower, engineer with Ciklum!

Interested already? We would love to get to know you! Submit your application. We can’t wait to see you at Ciklum.

#LI-AD1