Founding Machine Learning Engineer

Arcade is hiring a Founding Principal Machine Learning Engineer to build and productionize small, fast, on-prem, agentic tool recommendation and runtime ML models.

Responsibilities

• Run Arcade’s ML pipeline end to end, from data and training through evaluation and release
• Make model shipping routine by increasing reliability across training, eval, and deployment
• Fine-tune and train models for tool selection, routing, retrieval, and agent memory
• Use the right techniques for the job, including distillation, embeddings, rerankers, and more
• Expand into new territory starting with search and recommendation across tools and agent context
• Build and operate an evaluation system:
  • Offline evals against real agent traces
  • Online measurement in production
  • Head-to-head comparisons with Claude, GPT, and Gemini
• Close the loop from production agent traces and tool-call data to improved models, within enterprise data boundaries
• Quantize, optimize, and package models for customer VPCs and air-gapped environments; coordinate with the Runtime team on serving
• Choose the ML stack and write the strategy; help determine who to hire next in ML
• Apply expertise in modern agent systems (harnesses, memory, skills) to identify where models meaningfully improve agent capabilities and to call early if an approach won’t work
• Drive iteration speed: projects that take a week should take a day
• Own the model pipeline lifecycle from data ingestion to model publishing, including models across:
  • embedding
  • reranking
  • classification
  • recognition
  • other agent-related tasks
• Stand by deployed models backed by able data to cover outcomes; be willing to admit faults and review issues (including row-by-row spreadsheet work)
• Ensure everything runs in customer environments: their cloud, their hardware, and sometimes air-gapped networks
• Provide stable-but-young pipeline improvements based on insight and experience
• Lead a group making agent capabilities, not only API calls, with intelligent and efficient abilities that outpace competitors
• Own the models on the runtime path of every agent call served
• Productionize the first agent recommendation model and set the patterns for future work
• Own build-buy decisions for the stack with a healthy budget to spend
• Report directly to the Head of Engineering

Requirements

• 7+ years of software engineering experience; 4+ years training and shipping production ML systems
• Experience training or fine-tuning models that reached production and moved metrics customers cared about
• Expertise in production agent systems: harnesses, memory, skills, tool use, and sub-agents
• Know how and when fine-tuning works and how to apply it effectively
• Defensible evals with statistics fluency to determine whether deltas are real or noise
• Deployed models under real constraints: tight latency budgets, limited GPUs, or third-party infrastructure (vLLM, ONNX, TensorRT, llama.cpp, or equivalent)
• Strong Python for training and ML work; TypeScript or Go for production services that serve models
• Ability to pick the stack, document decisions, and defend them over time
• Preference for shipping a working v0.5 over a polished v2.0 on a later quarter timeline
• Comfort with ambiguity in an early team environment where the charter expands and decisions are made with incomplete data
• Strong drive to ship

Technologies

• Python, TypeScript, Go
• vLLM, ONNX, TensorRT, llama.cpp
• Claude, GPT, Gemini
• MCP

Compensation and Benefits

• Competitive salary, equity, and benefits
• $230,000 per year starting base salary (range aligned; determined by background, experience, and performance)
• In person at the San Francisco office

Location

• San Francisco, CA (onsite)

Traction

• Real deployments with Fortune-100 customers including Morgan Stanley and Open Table

Bonus Points

• Shipped ML into enterprise on-prem or regulated environments (financial services, healthcare, government)
• Tool-use benchmark or eval experience: BFCL, τ-bench, ToolBench, MCP evals, or equivalent
• Familiarity with the MCP ecosystem (extra bonus for filing an issue against the spec)
• Built training-data pipelines from production traces with privacy controls
• First ML hire experience at an early-stage company, with willingness to do some things differently
• Open-source contributions or published work that held up with other engineers
• Experience at an early-stage startup

Founder-Market Fit

• CEO previously founded Stormpath (acquired by Okta), creating the first Authentication API for developers
• CTO led the vector database team at Redis, shipped 100+ LLM applications, and contributes to LangChain and LlamaIndex

Dream Team

• Authentication, integrations, distributed systems, and AI experts from Okta, Redis, Microsoft, Splunk, Ngrok, Google, Airbyte, Disney, and HPE
• Team members have built and founded multiple successful developer platforms

Timing and Market

• Enterprise demand: almost none have agents in production yet
• Building critical infrastructure for a major technological shift

Backed By the Best

• Series A led by SYN Ventures with strategic investment from Morgan Stanley and Wipro
• Earlier investors include Databricks, Clickhouse, MongoDB, Perplexity, Cohere, ScaleAI, Confluent, Elastic, and Firebase