Job Title- ML Engineer – AI Suggestion & Data-Enrichment Systems
Experience- 4 to 6 Years relavent experience
Role:
Build ML components that power an enterprise tool where users assemble item/recipe-like
structures from a bespoke ERP catalogue. Your models will
- suggest the right items and alternatives,
- pre-populate grids/forms based on a brief,
- learn from user overrides, and
- turn unstructured inputs (notes, transcripts) into structured updates.
Everything has to work with role based workflows and sometimes incomplete ERP data.
Responsibilities
- AI-powered suggestions: Build ranking/recommendation models that propose the most
suitable items/components based on business variables such as cost, customer/profile,
category, service/class, availability, and preferred/platformed items.
- Auto-population of structures/grids: Train models to generate or pre-fill data grids/forms
from a short description or template (e.g. client type, duration, constraints) using items
already present in the ERP.
- Human-in-the-loop learning: Capture user actions (accept, reject, replace, mark-as
preferred) and feed them back to improve future suggestions for that unit/customer/profile.
- Use of complementary data: Join ERP data with extra attributes (nutrition/impact scores,
trends, satisfaction scores, custom business attributes) so ranking can optimize for more than one objective.
- Unstructured → structured (NLP): Take text or transcripts from review/presentation
sessions and extract structured changes (replace item A with B, adjust quantity, add
constraint) and map them to the right entities.
- Data pipelines & validation: Build reliable feature/data pipelines from ERP and external
sources; add checks for missing codes, outdated cost data, and duplicates so models don't
learn from bad rows.
- MLOps & serving: Package models behind low-latency services, register versions, enable
A/B or shadow runs, and monitor latency, coverage, and suggestion quality.
Must-Have Qualifications
- 4+ years in production ML (recommendation/ranking/search or adjacent).
- Strong Python and ML ecosystem (pandas, scikit-learn, plus PyTorch/TensorFlow).
- Experience combining rules/business constraints with learned models in one scoring pipeline.
- Solid NLP for classification/extraction; able to work from ASR/transcripts to structured fields.
- Comfortable with ERP/master-data–style inputs (incomplete, late, inconsistent) and building validation/normalization layers.
- Experience with MLOps (experiment tracking, model registry, CI/CD for ML, monitoring).
- Able to define and track acceptance rate, top-k hit rate, coverage, freshness as product ML metrics.
