Senior AI Architect

About Aivar

Aivar Innovations is revolutionizing enterprise AI transformation through production-grade accelerators and intelligent automation. We've delivered 60% efficiency gains, 95% faster processing, and 70% cost reductions for clients across regulated industries. Our AI-First philosophy powers flagship AI accelerators like Velogent AI (Agentic Process Automation), Convogent AI (Voice AI), and Kubogent AI (Enterprise AI Infrastructure on Kubernetes). We are looking for exceptional builders to join our journey.

What You'll Build

As Senior ML Architect, you'll architect the intelligence layer that powers Aivar's accelerators and client solutions. You'll design multi-agent systems for regulated industries, build production-grade ML infrastructure on Kubernetes, and pioneer agentic AI frameworks that autonomously execute complex business workflows. This isn't about optimizing accuracy by 2%—it's about building systems that eliminate 80% of manual work and unlock entirely new operational paradigms.

The Impact You'll Drive

Architect Production AI Systems

- Design end-to-end ML architectures for Velogent AI's framework, enabling governed multi-agent systems in financial services, healthcare, and life sciences
- Build scalable inference infrastructure leveraging for sub-second latency throughput
- Create MLOps pipelines integrating Kubeflow, MLflow, Ray, and Karpenter for distributed training and intelligent autoscaling on Amazon EKS

Accelerator Innovation & Client Delivery

- Lead technical design of AI accelerator features—from distributed compute orchestration to multi-tenant isolation and disaster recovery patterns
- Architect Convogent AI voice systems with real-time STT/TTS, contextual memory, and post-call analytics serving 250+ concurrent sessions
- Translate ambiguous business challenges into concrete ML solutions—from autonomous invoice processing to conversational intelligence

Technical Leadership & Evangelism

- Partner with customers and cloud teams on enterprise AI engagements, conducting architecture reviews and POCs for Fortune 500 clients
- Define best practices for LLM serving, prompt engineering, guardrails, and model evaluation across accelerator products
- Mentor ML engineers on agentic design patterns, production ML, and cloud-native AI infrastructure
- Contribute to internal R&D—experimenting with frontier models, emerging frameworks (MCPs, agent orchestration), and optimization techniques

What You Bring

Core Expertise

- 8+ years building production ML systems, with 3+ years architecting at scale (distributed training, high-throughput inference, or MLOps platforms)
- Deep hands-on experience with ML workloads— GPU scheduling, autoscaling (Karpenter), distributed frameworks (Ray, Kubeflow)
- Strong foundation in LLM architectures and Generative AI—prompt engineering, RAG pipelines, model serving (vLLM, Triton, Sage Maker), fine-tuning, and evaluation
- Proficiency in MLOps tooling—experiment tracking (MLflow/Weights & Biases), feature stores, model registries, CI/CD for ML, monitoring frameworks

Technical Depth

- Solid software engineering skills in Python and familiarity with ML frameworks (PyTorch, Tensor Flow, Hugging Face)

- Experience with cloud-native architectures on AWS—Sage Maker, Bedrock, EKS, Lambda, managed databases, networking, IAM
- Understanding of agentic AI patterns—multi-agent orchestration, tool use (function calling), reasoning frameworks, validation layers
- Knowledge of voice AI technologies (STT/TTS, WebRTC, real-time streaming) or governance frameworks for regulated AI is a strong plus

What Makes You Special

- Builder mindset—you prototype fast, iterate based on feedback, and obsess over production reliability
- Business acumen—you translate technical decisions into ROI narratives that resonate with CTOs and business leaders
- Curiosity—you stay current with AI research, experiment with new models, and bring fresh ideas to the team
- Communication excellence—you create compelling architecture diagrams, write clear design docs, and present confidently to clients

Why This Role is Different

Real Production Impact:
Your architectures won't gather dust in Jupyter notebooks—they'll…