Applied AI Engineer

Applied AI Engineer

Sigma360 is looking for an Applied AI Engineer to turn hard compliance problems into working AI systems. You'll train models, build agents, engineer prompts, and validate everything against real-world data. It's a notebooks-first role for a hands-on builder who's energized by ambiguity and obsessed with making AI actually perform on messy, adversarial data.

We're hiring exceptional engineers across the junior-to-senior spectrum; final compensation reflects experience and level.

About Sigma360

Sigma360 is an MIT-incubated, venture-backed, Series B data and analytics company helping financial institutions, fintechs, and governments manage entity risk. We turn the world's messy, fragmented data into clear answers — powering name screening, sanctions compliance, adverse media monitoring, KYC investigations, and risk research for some of the world's most demanding compliance teams.

Engineers own architecture, AI & data science own model quality, and everyone owns impact.

Why This Role Matters

Compliance AI is genuinely hard. Financial crime is adversarial — names, structures, and jurisdictions shift constantly. The data is messy, multilingual, and incomplete. The stakes are real: a missed sanctions hit is a regulatory event; a flood of false positives buries the analyst teams that rely on our product.

We're automating some of the most repetitive, highly regulated back-office work in global finance — and building the AI systems that make it possible:

• Agents that reason about entities across fragmented data sources
• Models that match names across dozens of languages and scripts
• Classifiers that separate relevant adverse media from noise
• Workflows that compress hours of analyst work into minutes

This isn't a "fine-tune an open-source model and call it a day" problem. It rewards creativity, scientific rigor, and a real tolerance for iteration. It's notebooks-first and research-oriented — upstream work, not production infrastructure. If you're energized by hard, ambiguous problems and don't need a detailed task list to make progress, you'll thrive here.

What You'll Do

Model development and AI system construction

• Build new AI agents and agentic workflows — design tool calls, orchestrate multi-step reasoning, validate outputs end to end
• Fine-tune and train language models, embedding models, and classifiers on proprietary compliance data
• Engineer, test, and iterate on prompts for LLM-powered features — systematically measuring output quality and failure modes
• Construct and curate high-quality training and evaluation datasets; ensure stratified, representative coverage
• Validate AI systems against real-world edge cases before handoff to production engineering

Data science and evaluation

• Design rigorous evaluation frameworks: define metrics, build holdout sets, measure precision/recall, identify distribution shift
• Run data exercises to test and tune AI/ML systems — measure catch rate and false positive rate on real entity data, diagnose failure modes, iterate on filter configurations and thresholds
• Answer analytical questions from product and business teams about AI feature performance, usage, and value

Tech Stack

• Primary environment: Databricks (notebooks, workflows, scheduled jobs)
• Languages: Python — pandas, PySpark, and standard ML libraries
• ML / AI: PyTorch, Hugging Face, LLM APIs, custom frameworks
• Data: Delta Lake / lakehouse; data shipped downstream to Postgres and Neo4j
• Infrastructure: AWS
• Downstream consumers: Golang backend, React frontend

What We're Looking For

Required

• Python fluency — you write clean, readable, maintainable code in notebooks every day; you're comfortable with pandas/pyspark and large DataFrames
• ML fundamentals — you understand model training, evaluation, overfitting, precision/recall tradeoffs; you can design a proper eval from scratch
• LLM/AI systems experience — you have built something real with language models: an agent, a RAG pipeline, a fine-tuned classifier, or something novel
• Prompt engineering — you treat prompts like code; you test, measure, and iterate systematically rather than vibe-checking outputs
• Autonomy and intellectual curiosity — you're energized by ambiguous problems; you make progress without a detailed task list
• Bachelor's degree (or higher) in Computer Science, Machine Learning, Data Science, Statistics — or equivalent practical experience

Nice to Have

• Databricks experience (notebooks, workflows, Delta)
• NLP / NER background — multilingual entity recognition, text classification, information extraction
• Experience training or fine-tuning embedding models
• Agentic workflow design and debugging
• AML, sanctions, KYC, or financial crime domain knowledge
• Model risk management experience or familiarity with SR 11-7 standards
• MLOps basics — experiment tracking, model registry, evaluation pipelines

What We Offer

• Remote-first team with high autonomy and ownership — your work is visible and consequential
• Competitive compensation and meaningful equity in a Series B company growing rapidly
• Health, dental, vision, and other benefits (or local equivalent)
• Generous time off and a culture that supports continued learning
• Small team environment: you'll have real ownership and real influence from day one

How to Apply

Apply with your résumé and a short note on the hardest AI system you've built — what made it hard, and what you'd do differently today.

Sigma360 is an equal opportunity employer. We are committed to fair hiring practices and to creating a welcoming environment for all team members. All qualified applicants will receive consideration without regard to race, color, religion, gender, gender identity or expression, sexual orientation, national origin, disability, age, familial status, or veteran status.