About TradeTrax TradeTrax is a jobsite intelligence platform for production homebuilders. We capture real-time field data from builders and trades and turn it into the insights that drive cycle time compression, trade accountability, and operational performance. Our data is changing how the residential construction industry makes decisions.

Summary

We're hiring a senior software engineer to build and run the production AI pipeline behind a new feature, and to grow into broader backend work over time. The defining challenge of this role is engineering around non-determinism. An LLM in the application path doesn't behave like a deterministic service. We need someone who treats that as an engineering problem, instrumenting the workflow, defining the right quality metrics, and calibrating relentlessly, rather than as a black box to be trusted or feared. We are early in putting LLMs into production. We're not looking for someone with all the answers about voice or ASR. We're looking for a strong engineer who can build the measurement and feedback loops that turn a non-deterministic component into a reliable product feature, and who raises the whole team's game by example. What you'll do

• Build and operate the production backend for the AI pipeline, including STT and LLM orchestration and the services around them, working within the existing backend team and its architecture.
• Build the instrumentation and evaluation layer: define quality metrics (edit rate, normalization drift, transcription accuracy, field-audio performance) and the regression/eval harness that lets us iterate on prompts and providers with confidence rather than vibes.
• Treat the LLM as a measured, calibrated dependency. Establish the feedback loops that let us optimize it over time.
• Build and operate scalable backend services on AWS (Lambda, API Gateway, Docker, CI/CD).
• Evaluate and tune STT and LLM providers against our own metrics.
• After this feature reaches production readiness, extend into broader backend platform engineering.
• Set the example for agentic, AI-assisted engineering on the team, and help us build that capability as a company.

Must have

• Strong backend engineering fundamentals: senior-level Python, FastAPI or equivalent, REST API and service design, distributed systems.
• Production experience running services on a major cloud. Solid CI/CD, Docker, and operational practices.
• Has taken a non-deterministic component, an LLM or a similar probabilistic system, into production and built the instrumentation and evaluation to make it behave reliably. This is the discipline the role is built around, so it's the one thing we won't compromise on.
• The range to learn unfamiliar domains fast. We'd rather hire a great backend engineer and have them learn the voice stack than hire a voice specialist who's a weaker engineer.

Strong plus

• Depth in LLM orchestration and evaluation/regression infrastructure. You've built these, not just used them.
• A Claude Code-centric (or comparable agentic) workflow, with a track record of influencing how a team works by example.
• Thought leadership on AI-in-the-loop engineering. Opinions formed from shipping, not slideware.

Nice to have

• Direct experience with ASR providers (Whisper, Deepgram, AssemblyAI) or audio pipelines. Useful, not required; learnable on the job.
• Mobile audio capture / mobile voice pipeline experience.
• Bilingual or multilingual NLP experience.
• Hosted LLM API experience (Bedrock, OpenAI, Anthropic, Vertex).
• MLOps and AI infrastructure monitoring.
• Comfort in TypeScript / Node.js environments (parts of our stack live here).