AI/ML Engineer II

• Design and implement end-to-end ML workflows supporting semantic search, classification, entity resolution, and retrieval.
• Build production-ready AI services with strong attention to reliability, testability, and maintainability.
• Develop and tune embedding pipelines, retrieval systems, and retrieval-augmented generation (RAG) components.
• Collaborate with data engineering and backend teams to integrate ML capabilities into scalable systems.
• Implement evaluation workflows tied to measurable mission performance (accuracy, latency, robustness).
• Support deployment, monitoring, and versioning of ML models as part of a disciplined MLOps lifecycle.
• Participate in architecture discussions and propose solutions aligned to platform constraints and mission needs.

• B.S. or M.S. in Computer Science, Engineering, or related technical field.
• 3-6 years of experience building applied ML systems or NLP workflows.
• Strong Python development skills with ability to write production-quality services.
• Experience training, tuning, evaluating, and deploying ML models.
• Familiarity with modern ML/NLP libraries (Transformers, spaCy, scikit-learn, PyTorch).
• Exposure to cloud environments and containerized deployment patterns.
• Strong communication skills and ability to collaborate across teams.

Additional Valuable Experience

• Experience with embeddings, vector search, RAG, and semantic retrieval systems.
• Familiarity with MLflow, DVC, Kubeflow, SageMaker, or similar tooling.
• Experience with graph-based retrieval, agentic systems, or tool-use architectures.
• Experience supporting defense, intelligence, ISR, or mission environments.

Pulled from the full job description 401(k) 401(k) matching Vision insurance Health savings account Dental insurance Flexible spending account Life insurance, People who do well here tend to approach problems similarly.

• They are comfortable operating without full clarity.
• They pay attention to what actually happens, not just what was intended.
• They are willing to be wrong, adjust quickly, and improve based on real feedback.
• They take responsibility for outcomes, not just implementation.

This is not a good fit for someone who needs tightly defined problems or prefers distance from how their work is used.

Security Clearance

Some roles require an active Secret, Top Secret, or Top Secret/SCI clearance. Where required, candidates must be eligible to obtain and maintain the appropriate clearance level.

U.S. citizenship is required for all positions. Torch.AI does not sponsor employment visas.

If you do not currently hold a clearance but are eligible, sponsorship may be available depending on role and mission requirements.

Work Location

Most roles are based at our headquarters in Leawood, KS. Some hybrid/remote positions across the Arlington, VA, Washington, DC, and Maryland (DMV) region are available. Limited travel (<10%) may be required for some roles.

Compensation, Benefits, Incentives

We offer competitive, performance-aligned compensation tied to technical depth, clearance level, and mission impact.

Total Rewards Include:

• Competitive base salary
• Quarterly performance bonuses
• Equity participation within the first 12 months
• Unlimited PTO + 11 paid company holidays
• Professional development in a high-growth, mission-driven environment
• Weekly in-office catering at HQ

Benefits:

• 401(k) plan (no current employer match, but under consideration)
• PPO, HSA, and TRICARE Supplement medical options
• Above-market HSA contributions
• HSA, FSA, and Dependent Care FSA options
• Dental and vision plans above national averages
• Employer-paid life insurance (1× salary)
• Employer-paid Short-Term and Long-Term Disability
• Voluntary Accident, Critical Illness, and Hospital Indemnity coverage
• Up to $300/month in tax-advantaged commuter benefits

Torch.AI is an Equal Opportunity / Affirmative Action Employer committed to building a team that reflects the mission we serve.

If you want to build AI systems that move from ingestion to actionable insight and know exactly why they produced the answer they did, we should talk.

The world's most consequential systems need the world's best builders. A new era is taking shape. AI is no longer confined to models or interfaces. It is becoming the foundation of how decisions are made across governments, industries, and real-world environments. What matters now is not just access to capability, but how it is applied, controlled, and sustained over time. Torch.AI builds a government-owned reasoning infrastructure which creates a Reasoning Layer to enable machine reasoning at scale, in environments where it is hardest to achieve and least tolerant of failure. This is not incremental software. It is long-term infrastructure designed to endure, integrate, and evolve alongside the systems it supports. Not another dashboard. Not another workflow app. Not another black-box model glued to a PowerPoint. The Reasoning Layer is a modular architecture where data is connected, governed, transformed, represented, fused, reasoned over, and delivered into mission applications and operational workflows. It does not replace systems of record. It enables them to function better together. The Reasoning Layer is comprised of: ORCUS (ingestion, orchestration, governance), NEXUS (semantic representation, vectorization), HALO (graph-based fusion and reasoning) and various product and capability components deployed for specific customer use cases. We are seeking candidates who approach problems creatively, are comfortable operating without full clarity, and take responsibility for outcomes, not just implementation. If you're driven to strengthen U.S. defense readiness and protect national interests, Torch.AI offers meaningful impact at national scale. What Makes Torch.AI Different Torch.AI was founded on a simple operational insight: better use of data leads to better decisions. As data volumes increased across commercial, enterprise, and national security environments, the limiting factor was not collection, storage, or user interface design. The missing layer was machine understanding. And an infrastructure that could operate and reason between layers, preserve context, reconcile meaning, and make data useful for decisions in real time. We believe the government must own its data and decision environment. In mission and operational environments, the layer where data becomes context, context informs models, models support decisions, and decisions shape action cannot be controlled entirely by private technology vendors. Ownership does not mean the government must build every component itself. It means the government maintains stewardship and authority over the mission and operational layer. Commercial software, models, and services can contribute to the environment, but they should not capture the mission. We are craftsmen. We build with intention. We ship with discipline. You'll collaborate with engineers, data experts, veterans, and mission practitioners. You'll own meaningful work, move quickly, and see your systems deployed in production, often within weeks. We are fast-paced, entrepreneurial, and mission-driven. Every day is a new puzzle.