AI/ML Engineer (Part-Time, Hybrid NLP + Classification)

Hi there :)

Thanks for checking in to find out about our open position. We´ll provide as much information as possible, but please feel free to reach us if you have further questions. We´ll be happy to see your application, even if there are skills you don't quite master!

About Us

At South Geeks, we engage top-performing Software Engineers, Security Experts, and Data Analysts from Latin America to join our clients' teams worldwide. For over 8 years, we've been helping future-shaping companies scale faster by curating world-class tech talent and building long-lasting, strategic partnerships. We pride ourselves on a people-centered culture that powers innovation, collaboration, and excellence.

About the Client

Our client is a Fortune 500 global energy company running a strategic initiative to migrate historical Plant Maintenance data into SAP S/4HANA through a governed, auditable web application integrated with SAP CPI. The work runs in 8 Agile-Scrum sprints over 16 weeks, fully remote, with a small senior team operating end to end.

About the Role

We are looking for a part-time AI/ML Engineer to design and build the AI component of the platform: a hybrid rule-based plus ML system that maps contractor free-text into a 4-tier SAP catalog hierarchy, flags anomalies, and learns from operator corrections. This is a 4-month engagement and your active participation spans SP0 through SP3 (W1 to W10), at roughly 6 to 10 hours per week during active phases.

Assignment Highlights

- 4-month engagement (16 weeks total; active in W1 to W10)

- 6 to 10 hours per week during active phases; 60 hours total

- 100% remote, Eastern Time overlap for gate and sprint reviews

- BYOD

Key Responsibilities

- Author the AI/ML Architecture Design Document at Gate 0 (W2). This is the W2 blocker for the whole project and must be readable by a non-technical PM.

- Recommend an accuracy threshold for the catalog mapping engine.

- Build the 4-tier catalog code mapping engine for SAP catalog types (Object Part B, Symptom C, Cause 5, Activity A).

- Implement the AI anomaly detection module.

- Design and implement the "Train Model" feedback loop where operator corrections retrain the model.

- Wrap the model as a microservice for the main app.

We recommend a hybrid architecture: a deterministic rule-based layer plus a supervised ML classifier with human-in-the-loop oversight. A pure rule-based fallback is acceptable.

What You Need to Succeed in This Role

- NLP / text classification in Python (scikit-learn or HuggingFace transformers).

- AI architecture documentation: a clean model design doc readable by non-technical reviewers.

- Rule-based NLP (keyword extraction, regex, scoring logic).

- ML classification pipeline (training data preparation, evaluation, tuning).

- Anomaly detection in tabular data.

- Python data science stack (pandas, numpy, scikit-learn, optionally spaCy).

- Model feedback loop design.

- API / microservice wrapping of an ML model.

Nice to have: precision/recall trade-off framing for catalog mapping; SAP catalog code domain knowledge.

Our Team

We strive to create an inspiring and growth-oriented environment where everyone feels valued, heard, and empowered. We promote both personal and professional development, with individualized support for your needs and concerns. We aim to build a space where everyone can thrive.

What We Offer

- Long-term projects

- 100% remote work

- Payment in USD

- Paid Time Off (PTO)

- Work from Home (WFH) & Training reimbursement

- English lessons

- Technical training

- Career coaching

This position is available for candidates based in LATAM.