Company: CYURÆ

Location: Gurugram, Haryana

Work Mode: Full-time, On-site

Experience: 5+ years

Reports to: CTO / Founding Team

About CYURÆCYURÆ is building an AI-powered fashion intelligence platform focused on personalised styling, visual understanding, recommendation systems and premium digital fashion experiences. We need a senior, hands-on AI professional who can lead model development, guide our existing AI Engineer and take AI systems from experimentation to production.

About the RoleWe are hiring a Senior AI/ML Engineer with 5+ years of practical experience in AI, machine learning and deep learning. You will own the design, fine-tuning, evaluation, deployment and continuous improvement of CYURÆ’s AI systems.

This is not a prompt-engineering-only role. You should understand the complete AI lifecycle:

Problem definition → data preparation → model selection → fine-tuning → evaluation → deployment → monitoring → improvement

You will work closely with the CTO, product team, fashion experts, backend developers, mobile developers and UI/UX designers.

Key ResponsibilitiesAI Architecture and Technical Ownership

- Design and own CYURÆ’s AI and machine-learning architecture.
- Convert product requirements into AI workflows, model pipelines, APIs and measurable evaluation criteria.
- Decide when to use hosted models, open-source models, fine-tuned models, recommendation algorithms, retrieval systems or deterministic rules.
- Build model-agnostic services so models and providers can be replaced without major application changes.
- Review AI code, experiments and model performance.
- Mentor the existing AI Engineer and support future AI hiring.

Model Fine-Tuning and Customisation
- Fine-tune language, vision, multimodal and embedding models for CYURÆ-specific use cases.
- Build and maintain training, validation, test and human-reviewed evaluation datasets.
- Work with LoRA, QLoRA, PEFT, adapter tuning, quantisation and knowledge distillation.
- Handle dataset cleaning, deduplication, annotation, augmentation, class balancing and quality checks.
- Track model versions, checkpoints, training parameters and experiment results.
- Reduce hallucination, inconsistency and bias using improved data, validation and guardrails.
- Optimise models for accuracy, latency, GPU memory, throughput and cost.

Multimodal AI and Computer Vision
- Build systems combining text, images and structured user or product data.
- Develop image classification, visual embeddings, semantic image search, garment attribute extraction, image similarity, object detection, segmentation, colour extraction and visual tagging.
- Evaluate and use CLIP, SigLIP, ViT, vision transformers and vision-language models.
- Build confidence scoring, fallback logic and human-review flows for low-confidence predictions.
- Compare third-party vision APIs with self-hosted or fine-tuned models based on quality, privacy, latency and cost.

Recommendation and Personalisation
- Design personalised recommendation, ranking and reranking systems.
- Build user, style, item and interaction representations using embeddings and structured features.
- Develop candidate generation, filtering,



compatibility scoring, ranking, diversity and feedback-learning logic.
- Handle cold-start users, recent products, sparse data and changing preferences.
- Track precision, recall, NDCG, acceptance rate, diversity and coverage.
- Work with fashion experts to convert qualitative judgement into structured labels and evaluation criteria.

LLM, RAG and Structured Workflows
- Build reliable LLM and multimodal workflows using commercial and open-source models.
- Develop retrieval-augmented generation using embeddings, vector search, metadata filters and hybrid retrieval.
- Implement structured outputs using JSON schemas, tool calling, validators, retries, confidence thresholds and fallback models.
- Version prompts, model configurations and evaluation datasets.
- Create automated and human-reviewed evaluation suites to prevent regressions.
- Keep model-generated outputs separate from deterministic product and business logic.

Production AI Engineering and MLOps
- Convert notebooks and proofs of concept into maintainable production services.
- Build model-serving APIs using Python and FastAPI.
- Containerise AI services with Docker and deploy them on AWS.
- Work with services such as Bedrock, SageMaker, ECS/EKS, S3, SQS, CloudWatch, PostgreSQL and Redis.
- Use Hugging Face, vLLM, TGI, Triton or equivalent inference frameworks.
- Build model and prompt versioning, rollback, monitoring and cost-tracking systems.
- Implement logging, tracing, latency monitoring, error handling, retries and asynchronous processing.
- Optimise inference through batching, caching, quantisation and model routing.
- Maintain clean code, tests, documentation and deployment runbooks.

Evaluation, Security and Privacy
- Define quality, latency, safety, reliability and cost thresholds before releasing models.
- Build test cases for standard, edge, multilingual, low-quality and adversarial inputs.
- Monitor model drift, data drift, quality degradation and inference costs.
- Protect user images, profile data and preferences during training and inference.
- Prevent prompt injection, data leakage, cross-user exposure and unauthorised retrieval.
- Maintain traceability for model version, prompt version, input source, output, latency and cost.
- Follow privacy-by-design, least-privilege and data-minimisation principles.

Required Qualifications
- 5+ years of professional experience in AI, ML, deep learning or applied data science.
- Strong production-level Python skills.
- Advanced practical experience with PyTorch and Hugging Face.
- Proven experience fine-tuning open-source language, vision, embedding or multimodal models.
- Hands-on knowledge of LoRA, QLoRA, PEFT, quantisation or similar techniques.
- Experience building repeatable training, validation and evaluation pipelines.
- Strong understanding of transformers,



embeddings, attention mechanisms and deep-learning fundamentals.
- Experience in at least two areas: LLMs, computer vision, multimodal AI, recommendation systems, personalisation or semantic search.
- Experience deploying AI systems into production and performing model evaluation and error analysis.
- Experience with FastAPI or another Python API framework.
- Working knowledge of PostgreSQL, Redis and vector search tools such as pgvector, Pinecone, Qdrant, Weaviate, Milvus or OpenSearch.
- Experience with Docker, CI/CD, Git and cloud infrastructure.
- Experience with AWS or an equivalent cloud platform.
- Ability to mentor engineers, review code and make architecture decisions.
- Comfortable working full-time from our Gurugram office.

Preferred Qualifications
- Experience in fashion-tech, retail, e-commerce or consumer applications.
- Experience building recommendation, ranking or visual-search systems at scale.
- Knowledge of CLIP, SigLIP, ViT, segmentation models, diffusion models or multimodal foundation models.
- Experience with MLflow, Weights & Biases, DVC or similar tools.
- Knowledge of vLLM, TGI, Triton, ONNX or TensorRT.
- Experience with multilingual AI, particularly English, Hindi or Hinglish.
- Experience in an early-stage or zero-to-one product environment.
- Open-source contributions, research publications or patents are a plus.

This Role Is Not Suitable For
- Candidates whose experience is limited to prompt writing or third-party AI API integration.
- Candidates with only coursework, certifications or notebook-level projects and no production deployment.
- Candidates unable to explain the dataset, architecture, metrics and business outcome of their AI projects.
- Candidates seeking only a management role without hands-on development.

What Success Looks LikeDuring the first three to six months, you should be able to:
- Review and improve the current AI architecture.
- Establish model, prompt, dataset and evaluation versioning.
- Define measurable baselines for quality, latency, reliability and cost.
- Complete at least one relevant model fine-tuning or adaptation project.
- Deploy production-ready AI APIs with monitoring and rollback support.
- Improve recommendation or visual-understanding quality using measurable results.
- Mentor the current AI Engineer and strengthen the team’s engineering practices.

What We Offer
- High ownership in CYURÆ’s founding AI team.
- Direct collaboration with the CTO, CEO and core product team.
- Opportunity to shape model strategy, architecture and product direction.
- Hands-on work across fine-tuning, multimodal AI, computer vision, recommendations and MLOps.
- Competitive compensation based on experience and demonstrated capability.

How to ApplySend your CV, LinkedIn profile, GitHub/GitLab profile and relevant project links to [email protected].

- Please include your current location, current and expected compensation, notice period, confirmation that you can work on-site in Gurugram, and a short description of one model you fine-tuned—including the base model, dataset, method, evaluation metric and production result.

📌 Senior AI/ML Engineer — Hands-on AI Lead (Gurugram)
🏢 Cyurae
📍 Gurugram