A
Applied AI Engineer
Azimuth Partners
5 hours ago
Full-time
On-site
Los Angeles, California, United States
Azimuth already runs on AI systems the firm owns and directs: research agents, revision loops, sourcing workflows, and an internal CRM that must stay a truthful mirror of relationships rather than a busy-looking database. We need a builder who will turn that into a durable internal platform: the custom AI stack, the CRM and its integrations, and the proprietary agents the team actually uses every day.
This role will encompass product engineering for a small, high-standard firm: reliable agents, clean data contracts, human-in-the-loop review where judgment belongs to people. You will work next to the operators who live in these tools, so the bar is production usefulness.
What you will do
Own Azimuth's internal AI platform end to end: agent runtimes, tool wiring, evaluation, logging, and the workflows that support research, harvest/revision loops, sourcing, and delivery.
Build and evolve the firm's CRM and related systems so they stay truthful and safe for handling of confidential client context.
Design and maintain proprietary agents the team runs in production (research, drafting under house style, ops helpers), with clear contracts, tests, and failure modes operators can trust.
Integrate the stack with the tools the firm already uses (Google Workspace, Drive comment loops, email/calendar where authorized, Git-backed skills and docs, ATS where relevant)
Enforce the firm's standard in software: refusal paths, source requirements, human gates on anything that could invent people or touch external systems unsafely.
Improve reliability and operability: evals, traces, dashboards that matter, docs another engineer or power user can run, and cost/latency discipline.
Partner with the delivery team on what to automate, what must stay human, and how new capabilities show up in real workflows rather than as unused prototypes.
What you bring
3 or more years building production software, with clear ownership of systems other people depended on.
Hands-on experience with LLM-backed applications in production: agents or tool-using workflows, retrieval, evaluation, and the unglamorous work of making them reliable.
Strong generalist engineering: Python is the default here; comfort with APIs, data models, auth, and shipping small internal products quickly.
Product judgment. You can talk to operators, cut scope, and choose the boring correct design over the impressive fragile one.
Clear written communication. You document systems so the firm is not dependent on your memory.
Nice to have
Self-hosted or tightly controlled model serving experience (vLLM, Ollama, or equivalent), or strong opinions about when local vs API is right.
Prior work on CRM, ATS, workflow engines, or other systems of record where bad writes are expensive.
Familiarity with consulting, recruiting, professional services, or other high-trust operational domains.
Experience with human-in-the-loop review UIs, background job systems, or multi-agent orchestration that had to survive real failure.
#J-18808-Ljbffr
This role will encompass product engineering for a small, high-standard firm: reliable agents, clean data contracts, human-in-the-loop review where judgment belongs to people. You will work next to the operators who live in these tools, so the bar is production usefulness.
What you will do
Own Azimuth's internal AI platform end to end: agent runtimes, tool wiring, evaluation, logging, and the workflows that support research, harvest/revision loops, sourcing, and delivery.
Build and evolve the firm's CRM and related systems so they stay truthful and safe for handling of confidential client context.
Design and maintain proprietary agents the team runs in production (research, drafting under house style, ops helpers), with clear contracts, tests, and failure modes operators can trust.
Integrate the stack with the tools the firm already uses (Google Workspace, Drive comment loops, email/calendar where authorized, Git-backed skills and docs, ATS where relevant)
Enforce the firm's standard in software: refusal paths, source requirements, human gates on anything that could invent people or touch external systems unsafely.
Improve reliability and operability: evals, traces, dashboards that matter, docs another engineer or power user can run, and cost/latency discipline.
Partner with the delivery team on what to automate, what must stay human, and how new capabilities show up in real workflows rather than as unused prototypes.
What you bring
3 or more years building production software, with clear ownership of systems other people depended on.
Hands-on experience with LLM-backed applications in production: agents or tool-using workflows, retrieval, evaluation, and the unglamorous work of making them reliable.
Strong generalist engineering: Python is the default here; comfort with APIs, data models, auth, and shipping small internal products quickly.
Product judgment. You can talk to operators, cut scope, and choose the boring correct design over the impressive fragile one.
Clear written communication. You document systems so the firm is not dependent on your memory.
Nice to have
Self-hosted or tightly controlled model serving experience (vLLM, Ollama, or equivalent), or strong opinions about when local vs API is right.
Prior work on CRM, ATS, workflow engines, or other systems of record where bad writes are expensive.
Familiarity with consulting, recruiting, professional services, or other high-trust operational domains.
Experience with human-in-the-loop review UIs, background job systems, or multi-agent orchestration that had to survive real failure.
#J-18808-Ljbffr