AI Engineer

What you will accomplish:

• Run the inference serving layer on our own GPU hardware: choose and tune the serving stack (vLLM, SGLang, TensorRT-LLM) for high throughput and low latency. 
• Optimize aggressively: tensor parallelism, quantization (FP8, AWQ, GPTQ), KV-cache and prefix caching, continuous batching, speculative decoding, concurrency tuning. 
• Serve multiple models and features off shared hardware: multi-LoRA, routing, and request scheduling that balances internal workloads against latency-sensitive product traffic. 

• Make our AI workloads efficient: improve latency, throughput, and GPU utilization so we get the most out of what we run. 
• Build the visibility: instrument performance and usage across our AI surfaces so there's clear data on how everything is running. 
• Surface the technical tradeoffs (performance, latency, efficiency) so the people making the calls have what they need to make them. 

• Ship the in-app agent layer that helps families coordinate: proactive nudges, smart suggestions, agents that summarize, draft, schedule, and act for busy parents. 
• Build the substrate underneath: tools, memory, orchestration, guardrails, and evaluation harnesses, integrated cleanly with production APIs alongside our architecture team. 
• Work in nimble pairs with feature owners, standing up whatever's needed to test an idea, including a vibe-coded UI when that's the fastest path to a real customer. Ship rough, learn fast, harden what works. 

Who you are:

• Technical and hands-on with infrastructure: you like running real systems on real hardware and keeping them fast and reliable. 
• A full-stack builder who wants the app layer too: you don't want to be boxed into infra. When a feature needs shipping, you want to pick it up and ship it, not just hand it off. 
• Performance-minded: you treat latency, throughput, and efficiency as things to engineer deliberately. 
• Rapid-prototyping and AI-first, with modern tooling (Claude Code, agent SDKs) part of your craft. 
• Motivated by work that matters. Families rely on these products during real moments in their lives. 

What you bring:

• 5+ years shipping production software, including meaningful applied AI or ML work. 
• Demonstrated experience running and optimizing self-hosted LLMs on dedicated multi-GPU hardware: a serving stack (vLLM, SGLang, or TensorRT-LLM) and the optimization that comes with it (tensor parallelism, quantization, batching, KV cache). 
• A track record of optimizing inference performance and efficiency (latency, throughput, GPU utilization). 
• Strong Python and engineering fundamentals, with the full-stack range to stand up a quick UI, and the genuine desire to work app-layer features and not only infra. 
• Hands-on with agent frameworks (Claude Agent SDK, LangGraph, or similar), LLM APIs, embeddings, and RAG. 
• Comfortable with AWS and the devops this role owns: Docker, CI/CD, monitoring, and observability. 
• Experience building internal tooling or platforms others depend on. Bonus for Slack apps, MCP, or agent orchestration at team scale. 

• Medical: In Tandem pays 100% of the premium for employees AND 99% for all additional family members 
• 401k: Up to a 4% match with immediate vesting   
• Paid leave for all new parents   
• Learning & Development stipend for employees   
• Paid Time Off: 11 Holidays + Winter Break (3 Days) + Volunteer Time Off (1 Day) + Floating Holiday (1 Day) 
• Personal Time Off: 15 days for 0-1 years of employment, 20 days 1-3 years of employment
• Supportive and flexible working environment – work from anywhere!