Edge ML / Embedded Engineer

Data Ideology•Pittsburgh, PA

5d•Remote

About The Position

We are seeking a specialized Edge ML / Embedded Engineer to join our team on a contract engagement at the intersection of constrained hardware and on-device machine learning. This is a discovery, architecture, and feasibility engagement — the primary output is a validated technical architecture and a constrained proof-of-concept demonstrator that shows the core concept works, not a production system. The right candidate thrives in ambiguous early-stage technical work, is energized by the challenge of making AI run on hardware that was never designed for it, and produces clear written findings when the answer is ‘it depends on specs we don’t have yet.’

Requirements

Bachelor’s degree in Computer Science, Computer Engineering, Electrical Engineering, or equivalent professional experience in embedded systems or edge computing.
5+ years of hands-on experience in embedded systems engineering, edge computing, or on-device machine learning, with demonstrated work on constrained hardware environments.
Expert-level proficiency with at least one edge ML inference framework: TensorFlow Lite, ONNX Runtime, llama.cpp, or equivalent. Experience optimizing and quantizing models for CPU-only inference is required.
Strong understanding of memory management, real-time data stream handling, and concurrent processing in resource-constrained environments. Experience with C++, Rust, or Python with tight memory management is strongly preferred.
Experience with embedded Linux or equivalent OS environments, including ARM-based processors, limited RAM, and environments without GPU availability.
Familiarity with real-time data ingestion from hardware interfaces or industrial systems — including serial protocols, message bus architectures, or event-driven pipelines at the edge.
Comfortable operating in a Phase 0 discovery and feasibility mode — producing assessment findings, ADRs, and a constrained demonstrator rather than production-ready software.
Strong written communication skills with the ability to document hardware constraint findings, framework evaluations, and architectural trade-offs in formats usable by both technical architects and client stakeholders.

Nice To Haves

AWS familiarity preferred, specifically IoT Greengrass as a candidate edge runtime and IoT Core for device-to-cloud connectivity. Hands-on implementation experience is not required but direct familiarity strengthens the candidate’s ability to evaluate candidate architectures.
Experience with voice-to-text or text-to-speech pipelines in offline or low-connectivity environments is a plus.
Experience working in consulting or client-facing project environments is preferred.

Responsibilities

Assess target edge hardware against the requirements of an on-device inference loop: evaluate processor architecture, available memory, OS and runtime environment, and whether candidate edge runtimes (such as IoT Greengrass or equivalent) can be supported.
Evaluate candidate edge inference frameworks for CPU-only SLM deployment — including TensorFlow Lite, ONNX Runtime, llama.cpp, and equivalents — assessing quantization approaches, inference latency, and memory footprint against feasibility targets confirmed during discovery.
Assess real-time data ingestion feasibility from operational subsystem interfaces, evaluating candidate patterns for consuming concurrent data streams within the memory and compute constraints of the target hardware.
Design and evaluate local data store options for the on-device SLM context, including storage formats, retrieval latency, and update mechanisms appropriate for the edge environment.
Build a constrained feasibility demonstrator on laptop or workstation hardware using simulated data feeds. The demonstrator validates the interaction model and core architectural approach — it is not a production prototype and does not connect to operational systems.
Implement a small number of scoped interaction flows in the demonstrator, integrating the voice interface pipeline with the SLM inference and local data retrieval components as agreed through the engagement scope.
Collaborate with the AI/ML Architect on SLM selection, domain restriction approach, and inference pipeline design — providing hardware and runtime constraint inputs that shape what is architecturally feasible.
Collaborate with the AWS Solutions Architect on the edge-to-cloud data channel, identifying what can realistically be buffered and transmitted from a constrained edge device under variable connectivity conditions.
Document hardware assessment findings, framework evaluations, and architectural trade-offs as Architecture Decision Records (ADRs) with explicit rationale. Clearly flag where recommendations are conditional on hardware or interface specifications not yet confirmed.
Communicate technical constraints and feasibility findings clearly to both technical architects and non-technical client stakeholders throughout the engagement.