Data Scientist- Hybrid (3 times per week)

Fusemachines•New York, NY

18h•$140,000 - $170,000•Hybrid

About The Position

Fusemachines is a global provider of enterprise AI products and services, founded in 2013 with the mission to democratize AI. Using proprietary AI Studio and AI Engines, the company assists clients in their AI Enterprise Transformation. With offices across North America, Asia, and Latin America, Fusemachines offers a range of enterprise AI solutions and specialized services enabling organizations of all sizes to implement and scale AI. They serve industries like retail, manufacturing, and government, and are dedicated to democratizing AI through high-quality AI education in underserved communities and helping organizations maximize their AI potential.

Requirements

3–8 years of experience in data science, machine learning engineering, or applied ML (mid-to-senior).
Strong Python skills for data analysis and modeling (pandas/numpy/scikit-learn or equivalent).
Strong SQL skills (joins, window functions, aggregation, performance awareness).
Solid foundation in statistics (hypothesis testing, uncertainty, bias/variance, sampling) and practical experimentation mindset.
Hands-on experience across multiple model types, including: Classification & regression, Time series forecasting, Clustering/segmentation.
Experience with deep learning in PyTorch or TensorFlow/Keras.
Strong problem-solving skills: ability to work with ambiguous goals and messy data.
Clear communication skills and ability to translate analysis into decisions.

Nice To Haves

Experience with Databricks for applied ML (e.g., Spark, Delta Lake, MLflow, Databricks Jobs/Workflows).
Experience deploying models to production (APIs, batch pipelines) and maintaining them over time (monitoring, retraining).
Experience with orchestration tools (Airflow, Prefect, Dagster) and modern data stacks (Snowflake/BigQuery/Redshift/Databricks).
Experience with cloud platforms (AWS/GCP/Azure/IBM) and containerization (Docker).
Experience with responsible AI and governance best practices (privacy/PII handling, auditability, access controls).
Consulting or client-facing delivery experience.
Causal inference experience (e.g., quasi-experimental methods, propensity scores, uplift/heterogeneous treatment effects, experimentation beyond A/B tests).
Agentic development experience: designing and evaluating agentic workflows (tool use, planning, memory/state, guardrails) and integrating them into products.
Deep familiarity with agentic coding tools and workflows for accelerated product development (e.g., AI-assisted IDEs, code agents, automated testing/refactoring, repo-aware assistants), including strong judgment on quality, security, and maintainability.

Responsibilities

Translate business questions into ML problem statements (classification, regression, time series forecasting, clustering, anomaly detection, recommendation, etc.).
Collaborate with stakeholders to define success metrics, evaluation plans, and practical constraints (latency, interpretability, cost, data availability).
Use SQL and Python to extract, join, and analyze data from relational databases and data warehouses.
Perform data profiling, missingness analysis, leakage checks, and exploratory analysis to guide modeling choices.
Build robust feature pipelines (aggregation, encoding, scaling, embeddings where appropriate) and document assumptions.
Train and tune supervised learning models for tabular data (e.g., logistic/linear models, tree-based methods, gradient boosting such as XGBoost/LightGBM/CatBoost, and neural nets for structured data).
Apply strong tabular modeling practices: handling missing data, categorical encoding, leakage prevention, class imbalance strategies, calibration, and robust cross-validation.
Build time series models (statistical and ML/DL approaches) and validate with proper backtesting.
Apply clustering and segmentation techniques (k-means, hierarchical, DBSCAN, Gaussian mixtures) and evaluate stability and usefulness.
Apply statistics in practice (hypothesis testing, confidence intervals, sampling, experiment design) to support inference and decision-making.
Build and train deep learning models using PyTorch or TensorFlow/Keras.
Use best practices for training deep learning models (regularization, calibration, class imbalance handling, reproducibility, sound train/val/test design).
Choose appropriate metrics (AUC/F1/PR, RMSE/MAE/MAPE, calibration, lift, and business KPIs) and create evaluation reports.
Perform error analysis and interpretation (feature importance/SHAP, cohort slicing) and iterate based on evidence.
Package models for deployment (batch scoring pipelines or real-time APIs) and collaborate with engineers on integration.
Implement practical MLOps: versioning, reproducible training, automated evaluation, monitoring for drift/performance, and retraining plans.
Communicate tradeoffs and recommendations clearly to technical and non-technical stakeholders.
Create documentation and lightweight demos that make results actionable.