AI/ML Engineer

About the Role

Lenskart is seeking a Senior ML Engineer to join its Data Science team. This team builds the recommendation engine that determines which frames millions of customers see across our app, website and a large network of physical stores.

Our recommendation service combines facial-compatibility scoring from computer vision, interaction-based mod- els on AWS Personalize, store-level inventory and bestseller signals, and an LLM re-ranking layer, all served through a low-latency API. The successful candidate will take end-to-end ownership of this system: improving ranking quality, strengthening the vision pipeline, applying Generative AI responsibly, and ensuring the platform remains fast and reliable at scale.

This is a high-visibility role with direct influence on customer experience and store conversion. It is well suited to an engineer who enjoys owning outcomes, working across model, API and infrastructure layers, and making decisions grounded in data.

Required Qualifications

• 4+ years of experience building and deploying ML-powered products in production, including at least 2 years on recommendation, ranking or search systems.
• Demonstrated end-to-end ownership of an ML system, from problem framing and data preparation through deploy- ment, monitoring and iteration.
• Strong command of recommender fundamentals: collaborative filtering, content-based and hybrid models, learning- to-rank, cold start, popularity bias, diversity and exploration.
• Ability to combine multiple signals, such as user behaviour, facial features, store inventory and bestseller data, into a single, well-calibrated ranking.
• Hands-on computer vision experience in face detection, landmarks or classification, using PyTorch or TensorFlow and OpenCV.
• Understanding of how image quality, lighting and device variation affect vision models in real-world conditions, with experience designing graceful fallbacks.
• Proven experience integrating LLMs into production systems, including prompt design, structured outputs, evalua- tion, and control of latency and cost.
• Solid AWS experience, ideally with AWS Personalize or SageMaker, along with containerized deployment using Docker and Kubernetes (EKS).
• Experience with CI/CD, autoscaling and safe release practices, along with sound security practices such as IAM roles and secrets management.
• Strong backend engineering skills in Python, with a track record of designing resilient, scalable REST APIs for high-traffic environments.
• Experience with timeouts, retries, circuit breakers, rate limiting, API versioning and graceful degradation when down- stream services fail.
• Working knowledge of Redis caching strategies, MongoDB, SQL and Elasticsearch.
• Comfort with metrics and experimentation, including Precision, Recall, NDCG, A/B testing and statistical significance, and the ability to translate findings into clear decisions.
• Bachelor’s or Master’s degree in Computer Science, Engineering, Statistics or a related field, or equivalent practical experience.

Preferred Qualifications

Candidates who bring experience in one or more of the following areas will stand out:

• Java and Spring Boot: hands-on experience building and maintaining production services in Java with Spring Boot, as our core recommendation service runs on this stack.
• Advanced ranking and exploration: applied experience with contextual bandits (Thompson Sampling, UCB) or reinforcement learning to balance exploitation with discovery, and with techniques that reduce popularity bias and improve catalog coverage.
• Embedding-based retrieval: building semantic or visual similarity search using vector databases such as FAISS, OpenSearch k-NN or pgvector.
• Model serving and optimization: deploying and tuning models with Triton, TorchServe, ONNX Runtime or Ten- sorRT, including quantization and GPU versus CPU cost trade-offs.
• Data and ML pipelines: experience with feature stores, workflow orchestration (Airflow, AWS Step Functions) and streaming platforms (Kafka, Kinesis) for near real-time signals.
• GenAI evaluation: designing offline and online evaluation frameworks for LLM outputs, including guardrails, human review loops and cost monitoring.
• Infrastructure and observability: infrastructure as code using Terraform or CloudFormation, and monitoring with Prometheus, Grafana or OpenTelemetry, including ownership of latency SLOs.
• Domain exposure: background in retail, omnichannel commerce, fashion or eyewear, particularly with store-level inventory and assortment challenges.