We need a full time engineer focusing on - Using existing AI APIs for labeling/data augmentation - Building data quality validation pipeline - Fine-tuning SLMs using the validated labeled data (LLM -> SLM distillation) - Models performance testing - Model packaging (Docker + Python) - Documentation
Job Requirements
We’re looking for an NLP / ML Engineer to own the full lifecycle of our small language model (SLM) pipeline — from data labeling & augmentation using existing AI APIs, through quality validation and fine-tuning, all the way to packaging models for production and documenting everything clearly.You’ll work closely with the founders and product team to turn messy, real-world text data into reliable, efficient, and well-tested SLMs that power our analytics platform.What You’ll DoModel & Data Pipeline WorkUse existing AI APIs (OpenAI, Google, etc.) for labeling and data augmentation.Design and implement data quality validation pipelines to ensure labeled data is consistent and trustworthy.Fine-tune Small Language Models (SLMs) using validated labeled data, including LLM → SLM distillation workflows.Build, run, and automate model performance tests (accuracy, robustness, latency, cost, etc.) on our NLP models.Productionization & ToolingPackage models and pipelines using Python + Docker for easy deployment.Write clean, modular, production-grade Python code (with tests where appropriate).Maintain clear documentation for datasets, training pipelines, models, and APIs so others can work with your outputs.Must-Have Skills & ExperienceCore Tools & EnvironmentDaily comfort in Google Colab or Jupyter notebooks.Advanced Python skills (clean code, debugging, profiling, virtual environments).Python Libraries (Mandatory)Hugging Face Transformersscikit-learn (sklearn)PyTorch (torch)google-generativeaiOpenAI Python SDKSLM / NLP FrameworksHands-on experience fine-tuning or using:BERT-based models (BERT, DistilBERT, DeBERTa, etc.)T5 (encoder/decoder) or similar sequence-to-sequence architecturesData & FormatsFluency in JSON and common data interchange formats (CSV, Parquet, etc.).Comfortable designing and working with schema for prompts, labels, metadata, and evaluation outputs.CommunicationStrong English (spoken and written) — you’ll read papers, discuss architecture, and document in English.Nice to HaveExperience with AWS services such as Lambda, ECS, EC2, S3.Practical experience with Docker (building images, pushing to registries, basic best practices).Working knowledge of SQL (writing queries, basic optimization, integrating ML pipelines with databases).What Success Looks Like in This RoleWithin the first 3–6 months, you will have:Designed and shipped at least one end-to-end pipeline: from labeling / augmentation via AI APIs → validation → SLM fine-tuning → evaluation.Established reusable templates for experiments, metrics tracking, and model comparison.Delivered Dockerized models with clear documentation that other engineers can deploy without you.”
