Applied A.I. / ML Engineer

ABOUT MANIFOLD

Strategy is the act of choosing what to do when you can’t know enough to be sure. Manifold builds software for the people making those decisions on behalf of the United States. The ones whose decisions shape the next decade, not the next quarter.

This work is still done largely by hand, by smart people with slide decks, spreadsheets, and instinct. We think it should be done with better tools. Not tools that make the decisions, but tools that sharpen the judgement of the people responsible for them. Cognitive enhancement, in the literal sense.

ABOUT THE ROLE

As an Applied ML Engineer at Manifold, you'll build forecasting models that DoD action officers use to make decisions months out, on data that's messy, partial, and rarely labeled the way you'd want. You'll own the work from model design through MLOps and production integration, pulling in techniques like causal inference and superforecasting where they actually move the needle. Your models ship, get used, and produce outcomes our users can measure.

This role demands more than algorithmic chops. You'll think in systems — how your pipeline lands inside a broader product, and how that product lands inside a partner's mission. And you'll write and explain what you've built to audiences from senior government stakeholders to the engineer two desks over, because in this domain, a model that can't be explained doesn't get trusted, and one that isn't trusted doesn't get used.

One honest note. This work is harder than most ML jobs. Ground truth is often sparse, so you'll make architectural calls without knowing if you were right for months. Explainability is non-negotiable. The interpretable model that builds trust usually beats the opaque one that doesn't. And the mission evolves; the problem you're hired against today may not be the one you're solving in 18 months. The right candidates treat all three as the reason to join, not reasons to hesitate.

WHAT YOU WILL DO

Deeply understand the mission before writing code: You will operate like our Forward Deployed Engineers, working with the customer to understand their mission and why their outcomes matter. You will use those insights to inform your architectural decisions.

Wrangle public and private datasources: You’ll partner with engineers to identify the right data for your model. When it doesn't exist, you will find substitutes or push back on the ask and propose something better.

Hypothesize, test, and compare model approaches: From causal inference and superforecasting to deep neural networks and Bayesian methods, you’ll compare and contrast different approaches to identify the right models for the mission. You will evaluate models with both quantitative rigor and the judgment to know when a metric is lying to you about whether the output is actually useful.

Configure and monitor end-to-end ML systems: You’ll stand up MLOps pipelines that catch drift and degradation before users do. You’ll architect end-to-end ML systems that don't just work in a notebook but hold up in production under real-world conditions. You will monitor models in production, partner with DevSecOps when something breaks, and work with the forward deployed team to decide how predictions surface to the end user.

Translate complex ML work into explanations that build trust with audiences: From senior government stakeholders to the engineer two desks over, you’ll verify that the people using your outputs actually understand them and can act on them. Explainability isn't a feature, it's the job.

WHAT WE VALUE

- Clarity of communication: You can explain what you built, how it works, and why it's trustworthy to a four-star general or a junior engineer with equal precision.

- Expertise in ML fundamentals and models: You've worked across model families: gradient boosting, deep learning, probabilistic methods, LLMs. And you know which earns its place for which problem.

- The taste to know when ML is not the answer: You know when to reach for a heuristic or simpler statistical model when that’s what the mission demands.

- An applied mindset: You ensure that every model you architect is tied to mission-critical outcomes. And you take ownership of ensuring the model goes from notebook to production.

- Judgment under ambiguity: The spec will rarely be clean. The right next step will rarely be obvious. You move anyway, with enough judgment to course-correct fast when you're wrong.

REQUIRED QUALIFICATIONS

- Bachelor's degree in computer science, electrical engineering, math, physics, or a related technical field OR equivalent experience shipping production machine learning systems

- Prior experience architecting end-to-end machine learning systems that have been deployed to an end customer

- Hands-on experience across multiple ML paradigms (e.g., classical ML, deep learning, probabilistic methods, GenAI)

PREFERRED QUALIFICATIONS

- Advanced degree in computer science, machine learning, artificial intelligence, or related technical field

- Prior experience building ML models in a startup environment

- Can point to writing, talks, or documentation that explains technical work to non-technical audiences

NICE TO HAVES

- Prior experience in defense

BENEFITS

- 401(k) with matching

- Comprehensive medical, dental, and vision coverage for you and your dependents

- Unlimited PTO

- Health & Wellness stipend

- Lunch stipend 3 times per week and dinner stipend after 8 p.m.

- Company-wide break the last two weeks of the year

- Supportive leave of absence including time off for military service, medical events, and parental leave