Location: fully remote

Start: ASAP

Long-term contract

Tourism Domain

Role Overview:

We are looking for A Machine Learning Engineer (Agentic AI) to own the full lifecycle of our proprietary Vision-Language Model (VLM) that powers a web-based agent for online bookings. You will train and deploy the model, build end-to-end data pipelines, and lead the RLHF process to ensure the agent generalizes across multiple booking platforms.

Key Responsibilities:

Train and deploy our proprietary Vision-Language Model (VLM) that powers a web agent for booking workflows.

Design, build, and maintain end-to-end data pipelines: data collection, cleaning, labeling, feature engineering, and model-ready datasets from several booking platforms and user interaction logs.

Own and manage the RLHF process:

• Define feedback schemas and annotation guidelines.
• Work with labelers/annotators to collect preference data and corrections.
• Train reward models and run policy optimization (e.g. PPO/DPO or similar).
• Continuously evaluate and improve model generalization across different booking platforms and UI variations (A/B tests, offline metrics, human eval).
• Work closely with product and engineering to translate business needs into ML requirements and ship reliable, user-facing features.
• Set up and maintain monitoring, experimentation, and observability for models in production (drift, quality, latency, failures).

Skills:

Must-have

• Proven hands-on experience training and deploying deep learning models (LLMs and/or VLMs) in production.
• Strong experience building end-to-end data pipelines in Python (data ingestion, transformation, labeling, storage; familiarity with workflow/orchestration tools is a plus).
• Practical RLHF experience: collecting human feedback, training reward models, and running RL-based fine-tuning (e.g. PPO, DPO, or similar methods).

Technical

• Strong proficiency in Python and modern ML stack (e.g. PyTorch /TensorFlow, Hugging Face ecosystem or similar).
• Solid understanding of deep learning for language and/or vision-language models: pretraining, finetuning, evaluation, and prompt engineering.
• Experience with MLOps practices: experiment tracking, reproducibility, model/version management, CI/CD for ML.
• Comfortable working with cloud environments and containers (Docker; any major cloud provider).
• Good knowledge of data structures, algorithms, and software engineering best practices (testing, code reviews, clean architecture).

Nice-to-have

• Experience with agentic frameworks (e.g. LangChain, LlamaIndex, custom tool-calling agents).
• Background in recommendation systems, ranking, or conversational agents.
• Experience in travel/booking/marketplace domains.