Full Stack Engineer; AI, Engineering

Through Line is the global safety net for people in crisis, helping millions access verified crisis support resources and guided self‑help on the world's largest platforms – ChatGPT, Google, Pinterest, Roblox, Tinder – just to name a few.

We're solving some of the hardest problems at the intersection of mental health, AI, and global scale. That means working alongside a team of engineers, clinicians and operators who care deeply about getting this right, where the stakes are real and the work matters.

We move fast and we trust our people. You'll have real autonomy and responsibility from day one. You'll be part of a team that's genuinely diverse because the people we serve are genuinely diverse. We're building Through Line to reflect that.

We're looking for a Full Stack Engineer—part hacker, part researcher, part builder. Someone who thrives in ambiguity and can span front‑end, back‑end, and AI systems. You want to set the global benchmark for AI safety in mental health.

Our AI care navigation platform guides users in distress through natural conversation, understanding their needs and connecting them to the right next step. It's embedded in customer products dealing with mental health, online safety, and crisis support.

We're moving from validated prototype to scalable infrastructure. You'll join at the point where AI safety engineering, product experimentation, and ethical system design converge.

We need someone who can navigate that intersection fluently—shipping fast without compromising on safety.

• Engineer for safety: design, test, and monitor guardrails for LLMs and classifiers.
• Instrument and evaluate model behaviour via our internal platform — your code helps train the next iteration.
• Work fluidly across research, front‑end, and back‑end (we value capability over titles).
• Ship features that matter. You'll see your code in production, helping people in distress every day.
• Stay close enough to the product to understand how safety decisions cascade through the system.

• Deep‑learning research. We're not publishing papers or running experiments for their own sake.
• Everything ships.
• Waiting for perfect specs. You'll work with incomplete information and incomplete specs. You iterate.
• Pure infrastructure work. You're not optimizing for elegance; you're optimizing for outcomes that matter.

• Versatile engineer
  . Comfortable shifting between research, API design, and UI; strong foundations in data structures, concurrency, and model behaviour.
• LLM-native. You integrate GPT, Gemini, Claude, or open models into your workflow; prompts and embeddings are second nature.
• Founder / CTO mindset
  . You’ve built or led something from scratch and owned it end‑to‑end.
• Bias toward action. You’ve migrated, switched careers, or pivoted under pressure — and thrived.
• Hacker mentality. You care more about outcomes than abstractions; you make it work, then make it safe.
• Mission-aligned. You believe technology should reduce suffering — not exploit it.

Our stack (for orientation):
React, Next.js, Type Script, Tailwind (frontend);
Node.js, Python (backend);
Google Cloud Run, Pub/Sub, Vercel (infra); GPT, Gemini, Claude, open‑weight models (AI)

• Vector search, embedding pipelines (Chroma, Pinecone, Weaviate)
• Agentic context‑engineering / LLM orchestration (Lang Graph, CrewAI, Transformers Agents, Auto Gen)
• Retrieval‑augmented generation / structured prompting / prompt‑chaining
• Multi‑model fine‑tuning (Open Pipe, Weights & Biases, Hugging Face)
• Synthetic‑data pipelines + LLM evaluation frameworks
• Secure dataflow / privacy‑by‑design in health or safety domains
• Early‑stage startup or social‑impact experience
• Open‑source or hackathon projects

• You’ve shipped a production‑grade LLM feature that improves safety or accuracy.
• You’ve improved or rebuilt a key workflow for faster iteration.
• Your development rules become part of our engineering culture.
• You’ve advanced our AI‑safety metrics — accuracy, latency, reliability.

The best benefit: your team

• A technical product leader who codes, with three…