Abdul Rehman

Lead applied AI developer

Hamburg European Open • Full time

Jan 2024 - Present • 2 yrs 4 mos

As a Generative AI Researcher & Developer at Project Hamburg Research, I worked on production AI systems, agentic workflows, RAG assistants, OCR/document intelligence, automation pipelines, and full-stack AI applications. I led architecture and delivery across MapleScholar, Local-MCP, and multiple production AI applications. My work included designing MCP-based tooling ecosystems with 25+ specialized tools across Python and Node.js services, integrating 6+ LLM providers, building fallback and caching mechanisms, and reducing API costs by 60%. I implemented LangGraph-based agentic workflows with structured outputs, JSON Schema/Pydantic validation, and production reliability patterns, achieving 95%+ response accuracy in real workflows. I also worked on tool orchestration, safe tool execution boundaries, retries, failure recovery, rate-limit handling, and prompt-injection-aware design. On the full-stack side, I delivered Next.js/Python applications with strong performance, including SSR/PWA systems optimized for sub-2.5s LCP, 90+ PageSpeed, and high-bandwidth production use. I also built multimodal OCR and document reconstruction systems that detect broken paragraphs, multi-page tables, interrupted layouts, and cross-page continuity issues. These systems used look-ahead logic and stateful cursor memory to preserve document structure for downstream search, QA, and automation. Additional work included multi-API research-data ingestion pipelines using OpenAlex, Semantic Scholar, arXiv, and Altmetric, plus AI enrichment, deduplication, and knowledge workflows. Beyond the CV-backed work, I also have hands-on experience with n8n automations, Zoom meeting creation, transcript ingestion, AI summarization, storage routing, AWS EC2/AMI migration, Lambda workflows, remote system replication, and self-hosted agent setups such as OpenClaw.