Data Scientist II

CommerceIQ builds AI software that large consumer brands use to manage how their products sell on retail platforms, covering content, media and sales, and says it serves more than 2,200 customers. The posting is pointed about the company's stance: it is not selling AI demos, it is shipping AI agents into Fortune 100 workflows every week. This role is applied machine learning against that commerce data, and the technical expectations lean heavily toward modern model adaptation rather than classical modelling alone. Two things are marked mandatory rather than preferred, which is unusual and tells you exactly what the interview will test: demonstrated success adapting foundation models to domain specific applications through fine tuning or transfer learning, and a strong ability to design, evaluate and improve models with robust validation strategies, bias and fairness checks and performance optimisation. At 3 or more years it is one of the more reachable data science roles in today's edition for someone who already has real modelling work behind them.

What the posting requires

- 3+ years of hands on experience in applied machine learning
- A strong background in machine learning, deep learning and natural language processing, with proven experience training and fine tuning large scale models such as language models, transformers and diffusion models
- Hands on expertise with parameter efficient fine tuning approaches: LoRA, prefix tuning, adapters and quantization aware training
- Proficiency in PyTorch, TensorFlow and the Hugging Face ecosystem
- Basic understanding of MLOps: experiment tracking, model versioning, CI/CD for ML pipelines and production deployment
- Experience with large datasets, feature engineering and data pipelines using tools such as Spark, Databricks or cloud native ML services including SageMaker, Vertex AI or Azure
- Knowledge of GPU and TPU optimisation, mixed precision training, and scaling workloads on cloud or HPC environments
- Excellent communication, able to explain complex ML concepts to technical and non technical audiences

Marked mandatory in the posting

- Demonstrated success adapting foundation models to domain specific applications through fine tuning or transfer learning
- Strong ability to design, evaluate and improve models using robust validation strategies, bias and fairness checks, and performance optimisation

Counts as a bonus, not a requirement

- Distributed training frameworks such as DeepSpeed, PyTorch Lightning or Ray
- Applied AI work across natural language, computer vision or multimodal systems
- Proven ability to lead and mentor junior applied scientists and ML engineers

Location and working pattern

Bengaluru.

A good fit if

You have a few years of real modelling work and want problems where the business outcome is measurable, since commerce data gives fast, unambiguous feedback on whether a model is helping.

Think twice if

Your fine tuning experience is thin. The posting marks foundation model adaptation as mandatory rather than preferred, so it is likely to be a hard filter rather than something you can compensate for elsewhere.