Artificial Intelligence Engineer

About Rebee.aiRebee.ai is an AI-powered resume parsing and hiring intelligence platform, built in Mumbai. We parse resumes into structured data, analyse job descriptions, and match candidates to roles semantically — by meaning, not keywords — across 17 industries and a library of 18,000+ tools and 65,000+ concepts.

We launched the full product in August 2026 and are now running pilots with early clients. Our models run in production on our own inference servers, so every millisecond and every point of accuracy matters.

Role overviewWe're hiring a Machine Learning Engineer to build, optimize and ship the models at the core of Rebee. This is a hands-on, model-first role: most of your time goes into creating ML models, making them faster and more accurate, and building the pipelines that train, evaluate and deploy them.

Location

Mumbai, India (on-site / hybrid)

Type

Full-time

Experience

2–5 years in applied ML / ML engineering

Reports to

Founder & CEO

You'll work on models such as named-entity recognition for resume parsing, embedding and reranking models for JD–resume matching, and classifiers for skills, roles and domains.

Key responsibilitiesModel development

• Design, train and fine-tune ML/NLP models — transformer-based NER, sentence embeddings, cross-encoder rerankers and few-shot classifiers.
• Own experiments end to end: data preparation, baselines, ablations and clear write-ups of what worked.
• Build and maintain evaluation sets and metrics (F1, precision/recall, NDCG, MRR) that reflect real hiring outcomes.Model optimization
• Cut inference latency and memory through quantization (INT8/FP16), distillation, pruning and ONNX / TensorRT export.
• Profile models on CPU and GPU; tune batching, sequence length and threading for throughput on our own servers.
• Shrink model images and containers without losing accuracy.ML engineering pipelines
• Build reproducible pipelines for data ingestion, labelling, training, evaluation and deployment.
• Package models as Dockerised microservices with stable APIs, health checks and versioning.
• Set up experiment tracking, a model registry and monitoring for drift and accuracy in production.
• Support batch and real-time inference, including large asynchronous parsing jobs.Collaboration
• Work with backend engineers to integrate models into the product, and with the founder on what to build next.Requirements
• 2–5 years building and shipping ML models to production (internships with real deployments count).
• Strong Python; solid PyTorch and Hugging Face Transformers.
• Hands-on NLP: fine-tuning transformers (BERT, DeBERTa or similar), embeddings and text classification.
• Proven model optimization work — quantization, distillation, ONNX Runtime or TensorRT — with measured before/after results.
• Experience building training and inference pipelines, and serving models behind APIs (FastAPI or similar).
• Comfortable with Docker, Linux servers and Git.
• Sound grasp of ML fundamentals: evaluation design, overfitting, data leakage, error analysis.
• B.Tech / B.E. / M.Tech / MS in Computer Science, AI/ML or a related field, or equivalent experience.Nice to have
• Information retrieval and ranking: bi-encoders, cross-encoders, vector search, hybrid search.
• Few-shot / low-data methods such as SetFit, or GLiNER-style NER.
• MLOps tooling: MLflow or Weights & Biases, DVC, Airflow or Prefect.
• GPU serving with Triton or vLLM; LLM fine-tuning (LoRA / QLoRA).
• Cloud ML on GCP or AWS; MongoDB.
• Multilingual NLP, or experience in HR-tech, recruitment or document AI.
• Open-source contributions, Kaggle results or published work.What we offer
• Real ownership: your models go straight into a live product used by recruiters.
• Early-team impact — shape Rebee's ML stack, tooling and practices from the ground up.
• Work on hard, applied NLP problems with direct access to the founder.
• Competitive compensation and ESOPs.