Senior AI Engineer

We are bringing order to the chaos of the home services industry. Subcontractor Hub is a comprehensive, AI-powered operating system tailored for contractors in solar, roofing, and HVAC. By centralizing leads, contracts, project management, and finances, we eliminate administrative bottlenecks and help contractors save time, reduce costs, and scale efficiently.

Our mission is to empower home service professionals to focus on growth rather than paperwork. AI is at the core of how we deliver that efficiency — and we're looking for the right engineering leader to help us push the boundaries of what our platform can do.

We are seeking an exceptional Senior AI Engineer who can bridge the gap between applied research and robust production systems. In this role you won't just be tweaking models or building local demos; you will own end-to-end AI capabilities that our customers rely on every day.

Working closely with our AI research team in Sofia, you will take their modeling patterns and research artifacts and apply the operational rigor needed to ship them as highly reliable, customer-facing features. You will set the standard for our AI organization through the quality of your code, your evaluation rigor, and your commitment to system safety.

• End-to-end ownership: take AI features from problem framing and research prototypes through productization and post-launch monitoring.
• Build agentic workflows: develop and orchestrate advanced multi-agent systems and LLM-based applications using Lang Graph and Python.
• Model development: train, fine-tune, and evaluate models with PyTorch and Hugging Face, applying LoRA and PEFT techniques where appropriate.
• Design rigorous evaluations: build the evaluation harness alongside the feature — dataset curation, metric definition, LLM-as-judge approaches, and agent-trajectory analysis.
• Ensure production safety: implement operational guardrails for a multi-tenant SaaS environment, including prompt-injection defenses, PII handling, output validation, jailbreak prevention, and circuit breakers for agent loops.
• Technical leadership and communication: raise the engineering bar with clear, precise PR descriptions, technical proposals, and architecture notes a distributed team can rely on.

• Backend API:
  Python, FastAPI, Pydantic, SQLModel
• AI and agents:
  Lang Graph, Lang Chain, Pydantic AI
• ML and model development:
  PyTorch, Hugging Face, scikit-learn, XGBoost, Neural Prophet
• Data processing:
  Polars, Pandas, Databricks
• PDF and browser automation:
  PyPDF, PyPDFium2, PyMuPDF, Playwright
• Databases and cache:
  PostgreSQL, MariaDB, Valkey
• Observability:
  Logfire, Sentry, Datadog

• Proven production experience: 5+ years of hands-on applied AI/ML work. You have shipped systems with real users behind them, not just research papers or personal projects.
• Agentic expertise: deep familiarity with prompt design, tool use, retrieval-augmented generation (RAG), memory architectures, and the failure modes that only appear at scale.
• Applied fine-tuning discipline: the practical knowledge to curate datasets and design evaluations, plus the discipline to know when fine-tuning is necessary versus when prompting or RAG is the right tool.
• Data fluency: comfortable writing SQL against large relational databases, building data pipelines, and extracting clear training signals from messy production data.
• Strong CS foundation: advanced proficiency in software development, pattern recognition, and neural networks.
• Education:
  
  master s degree or higher in computer science, AI, or a related field — or equivalent practical experience.

• Computer vision: experience with image segmentation, object detection, or polygon extraction is a strong advantage.
• Domain experience: previous work building AI applications for B2B SaaS, business management platforms, or the construction and contracting industry.