Do work that has to work in the real world.

What you will do

• Clean, explore, and preprocess structured and unstructured data to extract meaningful patterns and baseline metrics.
• Develop, evaluate, and benchmark machine learning and statistical models using sound validation techniques.
• Build and maintain reproducible data pipelines, ETL scripts, and analytical workflows for real-world datasets.
• Collaborate with engineering teams to package and serve model prototypes via lightweight APIs (FastAPI / Flask).
• Document analytical methodologies, dataset assumptions, validation results, and experiment findings with clarity.

What will help

• Proficiency in Python and core data libraries (Pandas, NumPy, Scikit-Learn)
• Strong foundation in statistics, hypothesis testing, probability, and exploratory data analysis
• Intermediate SQL for querying, aggregating, and structuring relational data
• Working knowledge of machine learning algorithms (regression, classification, clustering, ensemble methods)
• Familiarity with data visualization tools (Matplotlib, Seaborn, Plotly) and version control (Git)

Useful, not required: Experience with PyTorch, TensorFlow, NLP / LLM concepts, computer vision basics, vector databases, FastAPI, Docker, or PostgreSQL is useful.

What you will do

• Turn user and business needs into flows, wireframes, prototypes and polished interfaces.
• Plan and conduct lightweight research, usability reviews and design validation.
• Create and maintain responsive components, interaction patterns and design-system guidance.
• Work directly with engineers through implementation and visual quality review.
• Produce purposeful motion, explainers and product storytelling assets in After Effects.

What will help

• Intermediate-to-advanced Figma workflow and a strong case-study portfolio
• Interaction design, information architecture and visual hierarchy
• Responsive web and mobile product design
• Prototyping, accessibility and design-system fundamentals
• Working Adobe After Effects and motion-design proficiency

Useful, not required: Experience with Lottie, video editing, 3D, brand systems, usability testing or designing data-rich and AI-assisted products is helpful.

What you will do

• Lead discovery with users, stakeholders and domain experts.
• Define product outcomes, priorities, requirements and success measures.
• Maintain a clear roadmap while making disciplined scope and sequencing decisions.
• Work closely with design, engineering, data and go-to-market teams.
• Run release learning loops and communicate decisions with clarity.

What will help

• Practical product discovery and end-to-end delivery experience
• Strong written requirements, user stories and decision communication
• Working knowledge of analytics, experiments and product metrics
• Technical literacy across APIs, data and AI-enabled systems
• Stakeholder alignment, prioritization and release planning

Useful, not required: Experience with healthcare, B2B SaaS, workflow products, AI products, startups or regulated environments is valuable.

What you will do

• Design and implement reliable pipelines for model training, evaluation, and deployment.
• Build robust retrieval-augmented generation (RAG) frameworks with clean indexing and validation.
• Optimize model performance, memory footprint, and API latency for high-throughput client systems.
• Collaborate with product and data engineering teams to transition ideas into scalable microservices.
• Establish best practices in data versioning, experiment tracking, and model monitoring.

What will help

• Strong Python proficiency and deep understanding of PyTorch or TensorFlow
• Experience with LLM orchestration frameworks (LangChain, LlamaIndex) and API integrations
• Hands-on vector search engineering (pgvector, Qdrant, Pinecone)
• Applied MLOps fundamentals: Docker, container orchestration, MLflow, and cloud deployments
• Deep understanding of statistical testing, evaluation metrics, and error analysis

Useful, not required: Experience in healthcare data standards, Edge AI, tensor optimization (TensorRT), or custom fine-tuning of open-weight models is a strong plus.