TixelJobs
B
Bhriguvia muse

AI Architect

REMOTEPosted 4d ago
ML EngineerLeadFull-time#remote

About the Role

**ABOUT US**

We are a **US\-based startup building next\-generation software platforms for regulated life sciences manufacturing and laboratories**. Our products operate in **GxP\-regulated environments**, requiring strong engineering discipline, validation readiness, and security\-by\-design.

Here, you won’t just be a cog in the machine — you’ll be part of building the machine. If you thrive in environments where you can **wear multiple hats**, **roll up your sleeves**, and **see your work through from start to finish**, this is your opportunity to make a real impact.

We’re creating a culture of builders, thinkers, and problem\-solvers who aren’t afraid to challenge the status quo. Your voice will be heard, your ideas will matter, and your contributions will shape the future of our product and the industry.

We’re growing fast, and looking for a hands\-on Full Stack Development Lead – Life Sciences who’s ready to wear multiple hats and shape the future of digital pharma.

**WHAT YOU WILL DO**

As AI Architect, you will own the end\-to\-end design and delivery of the platform's AI/ML capabilities — from OT edge inference on constrained hardware to cloud\-scale LLM\-augmented workflows, computer vision for shopfloor quality, and an AI governance layer that satisfies FDA 21 CFR Part 11, EU Annex 11, and emerging AI\-in\-GxP guidance (GAMP 5, FDA AI/ML framework).

**DUTIES AND RESPONSIBILITIES:**

AI Platform Architecture

* Design a three\-horizon AI maturity roadmap: rule\-based decision support → supervised ML → autonomous orchestration
* Architect a tenant\-isolated AI inference layer using per\-tenant KMS CMKs and namespace\-level isolation
* Define the Unified Namespace (UNS) \+ MQTT Sparkplug B data spine for AI\-ready OT/IT convergence
* Lead selection and governance of LLM, embedding, and vision model stack (open\-weight vs. proprietary)

GxP AI Governance

* Design the AI Model Lifecycle Management (MLM) framework covering versioning, validation, drift monitoring, and retirement under 21 CFR Part 11 / Annex 11
* Author AI Risk Assessment documents aligned with GAMP 5 Category 4/5 and FDA AI/ML Action Plan
* Implement immutable audit trails for all AI\-generated records, predictions, and human\-override events
* Define Algorithm Change Protocol (ACP) procedures for post\-market model updates without full revalidation

LLM \& Intelligent Workflow

* Integrate LLM\-augmented workflows into various modules using modern AI \+ LiteLLM abstraction
* Build a retrieval\-augmented generation (RAG) pipeline with pgvector for regulatory needs,
* Design multi\-agent orchestration patterns for GxP\-compliant AI\-assisted review and approval workflows
* Implement prompt governance, hallucination detection, and output confidence scoring for regulated use cases

Computer Vision \& Edge AI

* Architect vision AI pipeline for Presence/Absence, Count, and Read detection in manufacturing environments
* Design edge inference stack
* Define synthetic dataset generation workflows for GxP\-compliant training data governance

Data \& MLOps

* Own the ML feature store design using TimescaleDB continuous aggregates
* Build MLOps pipeline: experiment tracking (MLflow/W\&B) → CI/CD model promotion → canary deployment → rollback
* Implement model observability: prediction drift, data quality, SLA alerting integrated with the platform's alarm subsystem

**WHAT YOU WILL BRING**

**Experience**

* ·Minimum of 8 years total experience

**REQUIRED TECHNICAL SKILLS**

* 8\+ years in software/data/ML engineering; 3\+ as architect
* Production LLM system design (RAG, agents, fine\-tuning)
* MLOps: experiment tracking, model registry, CI/CD pipelines
* Distributed systems: Kafka, event\-driven architecture, EKS
* Time\-series data: TimescaleDB, InfluxDB, feature engineering
* Computer vision: YOLO, SAM, CLIP/SigLIP, or equivalent
* Edge inference: TFLite, ONNX Runtime, quantization
* IaC: Terraform; cloud: AWS or Azure at scale

**REQUIRED DOMAIN \& REGUL
Share