ML / AI ENGINEER

ABOUT SKETCHDECK.AI

SketchDeck.ai is transforming construction estimation through applied AI. Our platform automates complex parts of the estimation process, helping construction professionals work faster, improve accuracy, and bid on more opportunities.

We apply computer vision and machine learning to real construction drawings, turning large, complex documents into structured information that estimators can review, correct, and use in their day-to-day workflows.

Our consistent record of delivering practical AI solutions has established SketchDeck.ai as a trusted technology partner within the construction industry.

ABOUT THE ROLE

We are looking for a Machine Learning / AI Engineer to design, train, evaluate, deploy, and improve the computer vision systems at the core of our products.

This is an applied engineering role with end-to-end ownership. You will work across dataset development, experimentation, model architecture, training, evaluation, inference, deployment, monitoring, and continuous improvement.

Our ML problems are grounded in real-world construction documents. Structural drawings contain dense visual information, inconsistent conventions, varying quality, complex geometry, text, symbols, and relationships between elements. Building reliable automation requires more than achieving a good benchmark score. Models need to perform consistently across customers, drawing styles, edge cases, and changing data.

You will work closely with product, full-stack, platform, and other ML engineers to turn model output into reliable production features. You will have direct visibility into how customers interact with your models and use that feedback to improve both model performance and the surrounding system.

This role is ideal for someone who enjoys the full applied ML lifecycle and wants their work measured by production outcomes.

WHAT SUCCESS LOOKS LIKE

During your first year, you will help us:

• Improve the accuracy, recall, precision, and robustness of our production computer vision models.

• Build repeatable training and evaluation pipelines that allow us to measure improvements with confidence.

• Improve dataset quality, labeling workflows, augmentation strategies, and management of difficult examples.

• Reduce the gap between offline model performance and real customer outcomes.

• Improve inference performance, reliability, and cost across production workloads.

• Establish stronger model versioning, deployment, validation, monitoring, and rollback practices.

• Build better diagnostics for understanding false positives, false negatives, regressions, and customer-specific failure patterns.

• Expand our use of OCR, geometric reasoning, document understanding, and other AI techniques where they create measurable product value.

• Improve the feedback loop between customer corrections and future model development.

• Help establish the engineering standards that support reliable ML development as our platform and datasets grow.

ABOUT YOU

• You care about solving real problems with machine learning, not using a particular model because it is new.

• You understand that a model performing well in a notebook is only the beginning. You care about datasets, inference, monitoring, failure modes, maintainability, and the customer experience around the model.

• You use evidence to guide model development. You can explain which metric matters, why it matters, and whether an improvement is statistically and operationally meaningful.

• You are comfortable investigating difficult model failures and working backward through data quality, preprocessing, training, architecture, inference, and application behavior.

• You understand the importance of reproducibility. You treat datasets, experiments, configurations, model artifacts, and evaluation results as engineering assets.

• You can work independently on ambiguous problems while collaborating closely with engineers and domain experts.

• You take ownership through production. You do not consider the work complete once a model artifact has been produced.

KEY RESPONSIBILITIES

COMPUTER VISION AND DEEP LEARNING

• Design, train, evaluate, and improve computer vision models for object detection, segmentation, classification, and related document understanding problems.

• Develop production models using PyTorch and modern computer vision architectures, including YOLO-family models and other architectures where appropriate.

• Improve model precision, recall, robustness, calibration, and generalization across different customers and drawing types.

• Investigate difficult examples and identify whether failures originate from the model, training data, labeling, preprocessing, post-processing, or surrounding product workflow.

• Design experiments that isolate variables and produce clear evidence for model and pipeline decisions.

• Evaluate new architectures, foundation models, and emerging AI techniques based on measurable improvements to product outcomes.

DATASET AND TRAINING PIPELINES

• Build and maintain pipelines for dataset creation, annotation, validation, preprocessing, augmentation, training, and evaluation.

• Develop strategies for identifying high-value examples for labeling and retraining.

• Improve training datasets through hard-negative mining, error analysis, representative sampling, and systematic treatment of edge cases.

• Establish clear dataset and model versioning so experiments can be reproduced and compared.

• Detect and prevent data leakage, poor dataset splits, annotation inconsistencies, and other issues that can produce misleading evaluation results.

• Partner with product and domain experts to translate customer corrections and production failures into better training data.

MODEL EVALUATION AND QUALITY

• Define evaluation frameworks that measure model performance against actual product requirements.

• Track metrics such as precision, recall, F1, confidence calibration, and class-level performance where appropriate.

• Build regression datasets representing important customers, drawing types, edge cases, and known failure modes.

• Develop visual debugging and comparison tools that make model behavior easy to inspect.

• Analyze model performance across different data segments rather than relying on aggregate metrics alone.

• Establish release criteria that prevent model improvements in one area from creating unacceptable regressions elsewhere.

PRODUCTION ML AND INFERENCE

• Design and maintain Python-based inference pipelines for batch and asynchronous production workloads.

• Optimize model inference for latency, throughput, GPU utilization, memory usage, and cost.

• Package and deploy models using Docker and cloud infrastructure across AWS and GCP.

• Design model-serving workflows that support versioning, validation, progressive rollout, and rollback.

• Build resilient inference processes that account for retries, duplicate work, timeouts, partial failures, and long-running jobs.

• Work with platform engineers to improve GPU scheduling, utilization, redundancy, monitoring, and recovery.

• Diagnose production issues across model serving, application services, queues, infrastructure, and data pipelines.

ML OBSERVABILITY AND CONTINUOUS IMPROVEMENT

• Build monitoring around model health, inference performance, failure rates, resource utilization, and production quality.

• Identify distribution changes and emerging failure patterns using production data and customer feedback.

• Maintain clear lineage between production outputs, model versions, configurations, and datasets.

• Build feedback loops that allow validated customer corrections to inform future model development.

• Help establish a disciplined process for model releases, regression testing, performance comparison, and production validation.

DOCUMENT AI AND GEOMETRIC REASONING

• Apply OpenCV and classical image-processing techniques where they provide simpler or more reliable solutions than deep learning alone.

• Work with OCR and document-understanding techniques to extract information from construction drawings.

• Develop solutions involving coordinate systems, transformations, spatial relationships, geometry, and structured visual information.

• Combine learned models, deterministic algorithms, and domain rules where hybrid approaches produce stronger production outcomes.

• Explore multimodal and foundation-model approaches where they can meaningfully improve extraction, validation, classification, or reasoning.

PRODUCT AND ENGINEERING COLLABORATION

• Work with full-stack engineers to integrate ML capabilities into customer-facing workflows.

• Define clear contracts between model inference services and application services.

• Partner with product and construction domain experts to understand how model errors affect real estimation workflows.

• Help design human-in-the-loop experiences where estimators can efficiently review and correct predictions.

• Contribute to architecture discussions, technical planning, code review, and engineering standards.

• Document experiments, architectural decisions, operational procedures, and important model behavior.

• Mentor other engineers and contribute to shared ML knowledge across the team.

MUST-HAVE QUALIFICATIONS

• 4 or more years of professional experience in applied machine learning, computer vision, deep learning, or a closely related field.

• Strong production experience building computer vision systems for object detection, segmentation, classification, or related problems.

• Strong PyTorch experience, including model development, training, evaluation, debugging, and optimization.

• Excellent Python skills and experience building maintainable production software, not only research notebooks.

• Strong understanding of computer vision fundamentals, including image preprocessing, augmentation, annotation, evaluation, and error analysis.

• Experience with OpenCV or comparable image-processing libraries.

• Strong understanding of deep learning architectures, training dynamics, loss functions, optimization, regularization, and model evaluation.

• Solid mathematical foundations in linear algebra, probability, statistics, optimization, and geometry.

• Experience developing repeatable training and evaluation pipelines.

• Experience deploying and operating machine learning models in production.

• Hands-on experience with Docker and cloud-based ML or GPU workloads.

• Experience optimizing inference performance and diagnosing GPU, memory, latency, or throughput constraints.

• Experience designing reliable evaluation datasets and identifying model regressions.

• Strong understanding of precision, recall, confidence thresholds, class imbalance, false positives, false negatives, and the tradeoffs between them.

• Experience working collaboratively with software engineers to integrate ML systems into production applications.

• Strong written and verbal communication skills.

• Master’s or PhD in Computer Science, Machine Learning, Computer Vision, Applied Mathematics, or a related discipline, or equivalent professional experience.

PREFERRED QUALIFICATIONS

• Experience with YOLO-family models or comparable modern detection architectures.

• Experience working with technical drawings, architectural plans, engineering documents, PDFs, maps, diagrams, or other visually dense documents.

• Experience with OCR, document understanding, layout analysis, or multimodal models.

• Experience working with geometric reasoning, coordinate transformations, spatial relationships, or 3D computer vision.

• Experience designing human-in-the-loop ML systems where users review or correct predictions.

• Experience turning production corrections and model failures into structured retraining workflows.

• Experience with experiment tracking, model registries, dataset versioning, and ML lifecycle tooling.

• Experience implementing model monitoring, drift detection, regression testing, and production quality metrics.

• Experience optimizing GPU inference workloads for performance and cost.

• Experience operating ML workloads across AWS or GCP.

• Experience with asynchronous job processing and distributed ML workloads.

• Experience working with MongoDB, object storage, and large image or document datasets.

• Familiarity with MLOps practices, CI/CD for machine learning, and automated model validation.

• Experience evaluating or integrating vision-language models, multimodal foundation models, or related modern AI systems.

• Experience in a growing technology company where ML engineers own work from experimentation through production.

• Regular use of AI-assisted engineering and research tools to accelerate development while maintaining rigorous review and validation.

OUR TECHNOLOGY

Machine Learning and Computer Vision

• Python

• PyTorch

• YOLO-family models

• OpenCV

• OCR and document-processing workflows

• GPU-based training and inference

Application and Data

• Python and Flask

• MongoDB and Amazon DocumentDB

• Redis

• Large PDF and image-processing workflows

Infrastructure and Operations

• Docker and Docker Compose

• AWS

• Google Cloud and GPU compute• Bitbucket Pipelines

• Sentry

• Prometheus

Our ML systems operate as part of a larger production platform used by construction professionals. You will work across the boundary between research, software engineering, infrastructure, and product to make those systems more accurate, reliable, and useful.

WHY SKETCHDECK.AI?

• Build AI systems that directly influence a real customer workflow.

• Work on complex computer vision problems using large, information-dense construction drawings.

• Own models from dataset and experimentation through deployment and production performance.

• See customer interaction and corrections inform future model development.

• Work directly with engineering leadership, product experts, and experienced software engineers.

• Influence our ML architecture, tooling, evaluation standards, and technical direction.

• Explore new AI capabilities where they solve measurable product problems.

• Join a team that values autonomy, accountability, practical engineering, and clear communication.

• Help build the AI foundation supporting the next stage of SketchDeck.ai’s growth.

Ready to build the AI behind the future of construction estimation?

Apply at careers@sketchdeck.ai