Backend Software Engineer - Applied ML & LLM Systems

About DwellyDwelly is a UK‑based, AI‑enabled lettings and property management platform that is growing through a roll‑up strategy acquiring estate agencies. The company leverages two arms: i) acquiring existing letting agencies and buying their highly sticky, recurring revenue‐type landlords portfolios, and ii) building a top‑notch technology to automate tenant management, payments and post‑rental property maintenance. Dwelly seamlessly integrates AI services to automate all business processes within brick‑and‑mortar real estate agencies, integrating them into a tech‑enabled digital letting platform in two months to radically improve the user experience and increase business efficiency.

Position Summary

We are looking for a Backend Engineer with strong applied ML experience to build production systems that extract, enrich, summarise and structure information from emails, documents and other unstructured data. This is not a pure data science or research role. It is a production engineering role focused on building reliable Python backend services around NLP, retrieval and LLM‑powered workflows. You will work on practical problems such as extracting useful information from email correspondence during agency migrations and summarising a client’s full communication history inside their Dwelly profile. The right person is comfortable working with messy real‑world data, taking prototypes into production, measuring quality and improving systems through evaluation and feedback loops.

What You’ll Do

• Build systems that extract structured data from emails, documents and other unstructured sources.
• Enrich migrated client, landlord, tenant and property records with useful information from communication history.
• Develop solutions that summarise a client’s full email history and surface the most relevant context inside Dwelly.
• Build production NLP/ML‑backed backend services that work reliably on messy real‑world data.
• Improve retrieval and ranking systems using approaches such as RAG, BM25, embeddings, hybrid search and reranking.
• Define quality metrics, evaluation datasets and feedback loops for extraction, summarisation and retrieval systems.
• Build Python backend services and APIs using frameworks such as FastAPI, Django, Flask or similar.
• Integrate ML and LLM workflows into production systems with clear error handling, observability and maintainability.
• Work closely with engineering, product and operations teams to turn real business problems into scalable automation systems.

What We’re Looking For

• Strong Python backend engineering experience.
• Experience with API frameworks such as FastAPI, Django, Flask or similar.
• Production experience with NLP, ML, information extraction, retrieval, ranking or summarisation systems.
• Ability to take research ideas or prototypes into production.
• Strong understanding of evaluation, metrics and quality measurement for ML/LLM systems.
• Practical experience with retrieval systems such as RAG, BM25, embeddings, hybrid search or reranking.
• Comfortable working with messy, ambiguous or incomplete real‑world data.
• Ability to build reliable services around ML workflows, including monitoring, testing and failure handling.
• Good understanding of LLM limitations, hallucination risks and safe user‑facing AI.
• Strong ownership mindset and ability to work independently in ambiguous product areas.

Nice to Have

• Experience building AI or LLM agents.
• Experience with document understanding, email parsing, entity extraction or CRM enrichment.
• Experience with LLM evaluation, prompt/version management or human‑in‑the‑loop review workflows.
• Experience with vector databases or search infrastructure.
• DevOps or CI/CD experience for deploying ML‑backed services.
• Experience testing ML systems on complex production datasets.
• Experience with typed programming languages such as TypeScript, Java, C#, C++, Kotlin, Scala or similar.

What Success Looks Like

• Useful information can be extracted from emails and documents with measurable quality.
• Client communication histories can be summarised safely, clearly and with relevant context.
• Retrieval and ranking systems improve over time through evaluation and feedback.
• ML and LLM workflows are reliable, observable and production‑ready.
• Operations and product teams can trust the outputs and understand when human review is needed.
• Unstructured data from acquired agencies becomes usable inside Dwelly faster and with less manual work.

Compensation & Benefits

Fully remote role.
Competitive compensation based on experience and impact.
Opportunity to work on high‑leverage automation systems at the intersection of backend engineering, applied ML, data and real operational workflows.
Competitive salary with the potential for equity options based on performan