Agents can now answer the phone, read documents, query operational data, and take actions inside a TMS. In freight, those actions move trucks and money. They must be correct, auditable, and safe.

About Alvys

Alvys is building the operating system for modern freight. Our AI-powered, cloud-native platform brings dispatch, driver mobile, safety and compliance, billing and settlements, native EDI, reporting, integrations, and automation into one product.

We were founded in 2020 by freight industry insiders who were tired of software built by people who had never moved a load. That combination of deep customer empathy and ambitious technology still defines us. We visit customers, learn how the work actually happens, prototype quickly, debate ideas openly, and turn the best ones into products people rely on.

More than 3,000 trucking and brokerage companies use Alvys today, including larger fleets with increasingly sophisticated operations. We've raised $77M to date, including a $40M Series B. That gives us real momentum and room to make bold bets, while remaining early enough that a great engineering leader can materially shape the product, team, and company.

We are building more than a modern interface for a traditional TMS. Foundry, our agent platform within Alvys, helps freight teams make decisions and complete work, and this team builds it. You'll work with curious, direct, low-ego people who solve hard problems, move quickly from prototype to production, and measure customer impact.

Our Principles

AI makes building cheap. It does not make judgment cheap.

The constraint has moved from writing code to deciding what to build and proving it is right. Agents belong in the middle of the pipeline. People remain accountable at both ends. These principles turn that into how we work every day.

Own the outcome

• You own what you ship. "AI wrote it" is not an answer.
• Judgment stays human. People decide what matters and whether the result is correct.

Verify independently

• Generation and verification are separate. Tests, review, and approval run independently of the agent that did the work.
• Autonomy is earned per workflow. Agents advance only on measured precision against human-verified outcomes. No number, no promotion.

Design for control

• AI-first means security-first. Customer data stays on approved infrastructure under managed identity.
• Cost is an engineering constraint. Every shared workflow has termination conditions and bounded fan-out.

Build durable systems

• Knowledge is tool-agnostic. Reviewed intent lives once; vendor-specific machinery is generated from it.
• Multiple vendors are a design assumption. Durable workflows don't couple to one model or harness.
• Twice is a pattern. Automate repeated work and share it within 48 hours.

The Data & AI Portfolio

The Data & AI portfolio covers Foundry, our embedded automation and agent platform, including a unified agent chat that answers questions and takes actions; Insights, conversational analytics grounded in each customer's own operating data; Intel, proprietary data intelligence drawn from third-party, licensed, and proprietary data sources; and Voice, AI voice and chat agents for load status, driver check calls, and the follow-ups in between. It also includes the shared foundation: the context graph, the data platform that feeds it, and the evals, tracing, guardrails, and model routing behind every agent.

Why This Role

Few engineering leadership roles combine frontier AI work, a proprietary dataset, and an industry this underserved. This one does.

• You build at the frontier, in production. Context graphs, self-learning agents, voice agents, and agentic automation are still being figured out across the industry. Here you ship them to real customers, on live freight operations, not benchmarks.
• You work with data no one else has. More than 3,000 trucking and brokerage companies run on Alvys, and more than $1 billion in funds moves through the platform every month. Combined with third-party and licensed sources, that data is the raw material for intelligence no single fleet or brokerage could build alone.
• Your work changes how an industry operates. With 91.5% of carriers running 10 or fewer trucks, most of our customers will never hire a data team. The agents you build give a small business the leverage of a much larger one.
• You own a portfolio, not a feature. Four products share one platform. You set its architecture, choose its models and vendors, and decide where the team invests as it grows.
• You shape the team from the start. The group is about eight engineers today. You'll hire the next wave and set the standard for how an AI-first engineering team works at Alvys.
• You stay close to the decisions. You report to the Head of Engineering, work directly with customers and our Forward Deployed Engineering team, and have a direct line into company planning, backed by a $40M Series B.

What You'll Be Responsible For

Leading and Growing the Team

• Manage, coach, and develop a team of about eight engineers, including senior engineers and tech leads, and grow it as the portfolio expands.
• Hire senior and staff engineers, and keep the quality bar high as the team scales.
• Run the team AI-first, with coding agents in the daily workflow and clear standards for how AI-written code is reviewed, tested, and shipped.

Owning Technical Strategy

• Set technical direction for agents, data, and the context graph, and make the build, buy, and model calls with cost, quality, and speed in view.
• Make tradeoffs across four products that share one platform, including what gets built, what gets cut, and where the team invests.

Building Agents That Learn

• Own the agent development lifecycle from prototype to production to continuous improvement.
• Build self-learning loops so agents improve from outcomes, user corrections, and transcripts, and make that improvement measurable.
• Make evals a release gate. No agent ships without tests for quality, safety, latency, and cost.

Data and the Context Graph

• Own the context graph that links loads, trips, carriers, drivers, customers, documents, and conversations, and the entity resolution that keeps it trustworthy.
• Partner on the data platform that feeds it, including event streams, CDC into Snowflake, data quality, and lineage, so agents and analytics share one source of truth.
• Capture decision traces (the inputs gathered, the rule applied, any exception, who approved it, and why) so precedent becomes searchable for agents and people, while rules stay explicit.
• Combine third-party, licensed, and proprietary data sources into intelligence customers can act on, with the provenance, licensing terms, and tenant boundaries of each source respected.

Safety, Isolation, and Reliability

• Design agents that change data with scoped permissions, human approvals, and audit trails, and defend against prompt injection and tool misuse.
• Enforce tenant isolation and least privilege across every agent, tool, and data path.
• Own reliability and unit economics. Track cost per task, per call, and per tenant, and keep margins healthy as usage grows.

Staying Hands-On

• Spend about a third of your time building: prototype, write evals, read traces, and debug production.
• Review your team's designs and code with real depth, and stay current on models and tools, separating what works in production from what demos well.

Cross-Functional Partnership

• Work as a peer to product and design leadership on direction, architecture, sequencing, and delivery.
• Work directly with customers and our Forward Deployed Engineering team, and turn repeated patterns into platform features.

The AI Fluency We Expect

We measure AI fluency by how you verify, not by how much you use AI. More autonomy is not always better. You match the level of autonomy to the risk of the work, and you hold yourself and the team to the same five competencies.

• Delegation: You decide whether to use AI at all, which task to hand off, and how much autonomy it gets.
• Description: You state goals, requirements, and limits clearly, from a single prompt up to project instructions, specs, and policy.
• Discernment: You check output against evidence: diffs, acceptance checks, evals, and production monitoring.
• Diligence: You own the effects, including data access, permissions, memory isolation, budgets, and kill switches.
• Systems: You know how models, context, tools, harnesses, and memory work, and where each one fails.

How You Work With Agents

• Design multi-agent systems on paper before you build them. For each step, choose a rule, a small model, a frontier model, or a person, and measure cost per outcome, not cost per token.
• Start from code-orchestrated workflows and add model-led autonomy only where the task needs exploration. Keep agent-to-agent handoffs few, and give each agent a contract: inputs, allowed tools, and output schema.
• Put limits in code, not in prompts: hooks, sandboxed worktrees, and a stop condition, retry cap, and budget on every loop.
• Separate the agent that writes from the agent that checks, and give the checker fresh context.
• Work in an explore, plan, execute, verify, and compound rhythm, where every correction becomes a rule, skill, hook, or test.
• Move fluently across Claude Code, Codex, and Cursor, and choose an editor-first or agent-first surface based on who should hold the controls for the next step.

Where You Take the Team

• Bring every engineer to running goal-based loops, where specs and checks are written first, never weakened to pass, and run by the agent itself.
• Move the team from steering every turn to designing the system that steers the agent: its context, its harness, and its loop.
• Build toward agents that run continuously in hosted sandboxes, each with its own least-privilege identity, a named owner, a budget, and a tested kill switch, with risk-scaled human review on financial actions, auth, settlements, and customer data.
• Close the improvement loop. Capture traces, run error analysis on real failures every week, encode the fixes, and gate every change to rules, memory, or evals behind evals and a human approver. Self-improving never means self-approving.

Minimum Qualifications

• 8+ years building software, including 3+ years managing engineers. You've grown senior engineers and tech leads and remain a hands-on builder who can prototype, debug production, and review code credibly.
• Production LLM systems. You've shipped LLM-powered products to real users, not demos, and can explain what broke, what you measured, and what you changed.
• Agentic engineering in your own work. You build with coding agents every day and can show the multi-agent workflows, contracts, hooks, and loops you put into production, and the checks that kept them honest.
• Depth in agent systems. You've built with tool use, function calling, MCP, orchestration, memory, and retrieval, and you understand where agents fail.
• Evaluation discipline. You've built eval systems that combine offline suites, online metrics, human review, and transcript analysis, and used them to gate releases.
• Strong data engineering judgment. You've worked with event streams, CDC, cloud warehouses, dbt, and data quality, and you know how the shape of the data limits what models can do.
• Distributed systems in production. You've designed and operated production services with real reliability requirements.
• Multi-tenant security. You've secured multi-tenant SaaS and isolated customer data, and you treat agent permissions as a security boundary.
• Language fluency. You write Python or TypeScript fluently and are comfortable reading C#, F#, and .NET. Experience with F# or another functional language is a plus.
• Startup-stage fit. You've operated at Series B/C velocity amid ambiguity, thin support, competing priorities, and a bias toward shipping and learning.
• Clear technical communication. You can explain a tradeoff to an engineer, a customer, and an executive.

Preferred Qualifications

• Knowledge graphs, context graphs, entity resolution, or GraphRAG in production.
• Real-time voice agents, including telephony, speech-to-text, text-to-speech, turn-taking, and latency budgets.
• Fine-tuning, reinforcement learning from feedback, or prompt optimization against eval sets.
• Embedded automation or workflow platforms such as Activepieces, n8n, or Temporal.
• SRE experience, or a record of running systems other teams depend on.
• Logistics or freight experience, or another vertical where data is deeply entangled with operations.
• Founder or early-stage engineering leadership experience.
• A history of hiring, growing, and retaining excellent engineers.

Built by people who've moved freight. Come build the intelligence layer with us.

Learn more at alvys.com