Senior Data Scientist / AI Engineer

Job description

We need a hands-on senior Datascience & AI engineer who can build deep analytics pipelines in Python and implement a GenAI Q&A layer over enterprise data. The work is highly technical: data wrangling, metric computation, anomaly detection/forecasting (light ML), retrieval-augmented generation (RAG), and local LLM inference using Llama + Ollama.

Responsibilities (technical)

• Build robust analytics code in Python using pandas/numpy to compute, validate, and reconcile KPIs (costing, margins, QBR metrics, operational metrics).
• Write efficient transformations (vectorization, memory optimization), and implement repeatable pipelines with tests and data validation.
• Develop SQL to extract/shape datasets from enterprise sources and/or a cloud data warehouse; optimize queries as needed.
• Implement a governed GenAI “ask the data” prototype:
• Use Llama-family models via Ollama (or llama.cpp/vLLM as needed)
• Build RAG over structured + semi-structured data (chunking, embeddings, retrieval, reranking)
• Produce structured outputs (tables/JSON) and drill-down-ready answers
• Add basic guardrails: grounded responses, citations/traceback to data, and safe handling of sensitive fields.
• Apply light-to-moderate ML where useful:
• anomaly detection (cost variances, outliers, feed failures)
• simple forecasting / trend analysis for key metrics
• model evaluation and error analysis
• Create reproducible experimentation and evaluation:
• test question sets for the LLM
• accuracy/groundedness checks
• latency profiling and performance tuning
• Package deliverables for deployment (Docker, config management), and produce technical documentation/runbooks.

Required skills & experience

• 7+ years hands-on in data science / analytics engineering / ML engineering (individual contributor).
• Expert in Python, especially:
• pandas, numpy
• data cleaning, joins/merges, windowed calculations, time-series handling
• performance optimization (vectorization, profiling, memory management)
• Strong SQL (complex joins, aggregates, window functions; tuning mindset).
• Solid fundamentals in statistics and ML:
• feature engineering basics, evaluation metrics, overfitting awareness
• scikit-learn (or equivalent) for quick modeling
• GenAI implementation experience:
• Llama models (or comparable open LLMs)
• Ollama for local inference (or similar)
• RAG frameworks (LangChain/LlamaIndex) or custom retrieval pipelines
• embeddings + vector stores (FAISS/pgvector/Weaviate/Pinecone)
• Good engineering habits:
• unit tests, data tests, logging, error handling
• Git, CI basics
• Docker and environment management

Nice-to-have

• Snowflake experience (or similar modern cloud data platform).
• dbt experience (modeling, tests, docs).
• Experience with enterprise “semantic layers” or metric definitions at scale.
• Experience building lightweight APIs (FastAPI) for analytics/LLM endpoints.
• Familiarity with security constraints (RBAC concepts, masking, audit logs).

Tools/stack (typical)

Python, pandas, numpy, SQL, scikit-learn, Jupyter, Git, Docker, FastAPI (optional), LangChain/LlamaIndex (optional), Ollama, Llama models, vector DB (FAISS/pgvector/Weaviate), cloud data warehouse (Snowflake or equivalent).