Job Summary
We are hiring an ML Engineer to build production agentic applications that automate complex geoscience workflows. You will design agent-based systems, integrate tools and external APIs, implement persistent state management, provide agents with the right context, and deliver well-tested services with lightweight web interfaces for operators and end users.
Beyond implementation, you will help shape what these applications do: which workflows to automate, where people stay in the loop, and how results are presented and verified. You will also be responsible for designing and executing LLM fine-tuning strategies when model adaptation is required.
Key Responsibilities
- Shape the application logic and user experience: define what a good result looks like, remove what does not help users decide, and make outputs easy to verify.
- Design agent workflows with branches, loops, sub-workflows, and explicit routing, choosing the simplest structure that solves the problem.
- Implement human-in-the-loop approvals, and decide which actions an agent may take on its own.
- Give agents the right context: tool-based search over files, logs, code, and databases, using text search, SQL, and semantic retrieval where appropriate.
- Ground answers in cited evidence, enforce tool permissions, and guard against prompt injection and unsafe actions.
- Integrate tools and external APIs. Handle concurrency, retries, and error paths.
- Add database-backed persistence. Ensure workflows can pause and resume reliably.
- Design and execute LLM fine-tuning strategies when needed: data curation, supervised fine-tuning with parameter-efficient methods such as LoRA, preference optimization (e.g., DPO), reinforcement learning methods such as GRPO with verifiable rewards, evaluation sets, and rollback criteria.
- Instrument agent runs with tracing and continuous evaluation, making decisions and actions auditable and detecting quality regressions early.
- Deliver production-ready services with automated tests and CI/CD, and expose APIs where other teams or systems need them.
- Collaborate with domain experts, product, and engineering. Present clear design docs and trade-offs.
- Leverage AI coding agents such as Claude Code or Codex effectively and responsibly: plan and direct their work, review and test what they produce, and stay accountable for every change that ships.
Qualifications:
Required:
- Bachelor’s degree or higher in Computer Science, Engineering, Physics, Mathematics, Geophysics, or a related field.
- Strong academic performance, including a minimum cumulative GPA of 3.5 on a 4.0 scale for all earned degrees. Equivalent academic grading systems will be considered for degrees obtained outside the U.S.
- Strong programming and software engineering fundamentals, including asynchronous and concurrent programming.
- Sound design and product judgment: you can explain why a system is built the way it is, what was left out, and how its quality is measured.
- Hands-on experience building multi-agent or graph-based LLM workflows, using established frameworks or custom implementations.
- Experience designing how agents find information: tool-based search over files, logs, and databases, and semantic retrieval with reranking when needed.
- Practical LLM fine-tuning experience: parameter-efficient methods such as LoRA, dataset preparation and quality assurance, offline and online evals; a working understanding of RL-based fine-tuning and reward design.
- Experience with tracing and evaluation of LLM applications.
- Experience building web services and APIs (e.g., FastAPI), and working with PostgreSQL, Docker, and Git.
- Ability to harness AI coding agents such as Claude Code or Codex: understand how they work, break down tasks, guide their execution, critically review their output, recognize failures, and take ownership of the resulting code.
Preferred:
- Experience designing complex agent systems, including task decomposition, parallel execution, result aggregation, and human-in-the-loop orchestration.
- Hands-on experience with LLM post-training techniques (e.g., DPO, GRPO), including reward design for verifiable outcomes and multi-step tool use.
- Experience running enterprise-scale applications in production, including retries, idempotency, timeouts, and rate limits.
- Prior work in workflow automation; exposure to geoscience or scientific computing is a plus.
- Frontend experience building lightweight operator dashboards.
Houston Benefits:
- 401k Retirement Plan - when you contribute up to 5%, not only will we match that, but we will also add an additional 2%
- Free on-site fitness centre offering group exercise classes
- Short-term disability and long-term disability coverage at no cost to you
- Parental Leave Program allowing mothers and fathers to take time off from work for childbirth, adoption and to bond with your new child
- Pet Insurance – options to meet every budget
- Wellness Program – resources provided to support your personal wellness and participation in activities can earn you $65 quarterly.
