We re seeking a hands-on Sr. Data Science Architect
who can lead the end-to-end modeling lifecycle
from problem framing and experiment design to production deployment and monitoring while setting up the technical architecture
for ML/GenAI and agentic systems. This is not
a data-engineering-heavy role; you ll partner with DE/Platform teams, but your center of gravity is modeling excellence, MLOps, and AI solution architecture
that moves business KPIs.
What you ll do
Strategy & Architecture (Data Science first)
Own the technical vision
for data-science initiatives; translate ambiguous business goals into modellable problems, KPIs
, and NFRs/SLOs
.
Define reference architectures
for classical ML, deep learning, and agentic GenAI
(RAG, tool-use, human-in-the-loop) including model registry, evaluation harness, safety/guardrails, and observability.
Make build vs. buy
and model/provider choices (OpenAI/Claude/Gemini vs open-source), including optimization strategies (INT8/4, AWQ/GPTQ, batching, caching).
DS Leadership & Experimentation
Lead problem decomposition
, feature strategy, experiment design (A/B, interleaving, offline/online eval)
, error analysis, and model iteration.
Guide teams across NLP, CV, speech, time series, recommendation, clustering/segmentation
, and causal/uplift where relevant.
Establish rigorous quality bars
: data & label quality checks, leakage prevention, reproducibility, and statistical validity.
Productionization & MLOps
Architect CI/CD for models
(unit/contract tests, drift checks, performance gates), model registry/versioning
, and safe rollouts
(shadow, canary, blue-green).
Design monitoring
for accuracy, drift, data integrity, latency, cost, and safety (toxicity, bias, hallucination); close the loop with automated retraining triggers where appropriate.
Orchestrate RAG
pipelines (chunking, embeddings, retrieval policies), agent planning/execution
, and feedback loops for continuous improvement.
Stakeholders & Enablement
Partner with product, strategy/innovation, design, and operations to align roadmaps; run architecture and model review
sessions with clear trade-offs.
Provide technical mentorship
to data scientists/ML engineers; codify patterns via playbooks, ADRs, and reference repos.
Collaborate with Ops/SRE to ensure solutions are operable
: runbooks, SLIs/SLOs, on-call, and cost controls.
Governance, Risk & Compliance
Embed model governance
: approvals, lineage, audit trails, PII handling, policy-as-code; support GDPR/ISO/SOC2 requirements.
Champion human oversight
for agentic systems with clear escalation and decision rights.
Must-have qualifications
14 20 years
delivering AI/ML in production, with 5+ years
in an architect/tech-lead capacity. Expert Python
and ML stack ( PyTorch
and/or TensorFlow
), plus strong SQL
and software engineering fundamentals (testing, packaging, profiling).
Proven record architecting scalable DS solutions
on AWS/Azure/GCP
; hands-on with Docker
and Kubernetes
(collaborating with platform teams rather than building infra from scratch).
MLOps proficiency: MLflow/Kubeflow
, model registry, pipelines (Airflow / Prefect / Vertex / Bedrock / SageMaker pipelines), feature stores, and real-time/batch serving ( KServe/Seldon/Triton/vLLM/Ray Serve
).
Depth across traditional ML
and DL
(NLP, CV, speech, time-series, recommendation, clustering/segmentation) and the ability to select/prioritize the right approach for the KPI.
Excellence in communication
and stakeholder leadership
; experience guiding cross-functional teams (DS, MLE, DE, Product, Ops) to ship value.
Preferred qualifications
Agentic AI & RAG:
LangChain/LangGraph or equivalent orchestration; vector DBs ( pgvector
, Pinecone, Weaviate, Qdrant); retrieval policy design and evaluation.
Evaluation & Safety:
offline metrics (precision/recall, ROC/PR, BERT-F1, BLEU/ROUGE), LLM eval harnesses
, red-teaming, prompt/response guardrails.
Experimentation:
online testing at scale, counterfactual/causal inference, telemetry design.
Performance & Cost:
quantization, speculative decoding, KV caching, batching/collation, throughput tuning on CPU/GPU. Familiarity with data-viz/decision support
(Tableau/Power BI/D3) and UX/HCI
collaboration for human-in-the-loop designs.
Consulting experience or multi-vendor delivery; pre-sales/SoW exposure.
