AI Architect (Deep Learning)

We are hiring this position to own the design, training, and validation of our core model - and to bring this work fully in-house. Your mandate is to take technical ownership, validate, progressively in-source model development. This is a senior, deeply technical, hands-on role. You will set the direction for our deep learning stack, write and review code, and be accountable for whether the model hits its performance targets.

You do not necessarily need a medical or clinical background. Your job is to build and train the model that produces output for our clinical teams. You will report to the CTO and work alongside a dedicated Project Manager who handles timelines, vendor logistics, and coordination, so you can focus on the architecture and the science.

Scope. The scope is broad because the role is founding - but it grows in phases, not all at once:

• First ~6 months: Take ownership of the existing model and pipeline. Validate vendor deliverables against held-out test sets you control, establish a rigorous validation methodology, and drive the core model to its performance targets.
• Next 6-12 months: Insource training end to end, harden the training, round out infrastructure, and produce the technical documentation feeding our regulatory pathway.
• As we grow: Define, hire, and mentor the small ML engineering team that scales this work.

.What You Will Own

• Model architecture and training: a multimodal ensemble
• Vendor validation: technical evaluation of external model-training and deliverables, validated against held-out test sets you control
• Insourcing: the transition from vendor-led to in-house model development, including defining the ML engineering roles you will eventually hire and mentor
• Validation methodology: patient-level train/validation/test splits that prevent data leakage, per-modality performance breakdowns, and rigorous reporting
• Explainability: outputs for the clinical team to review to confirm the model attends
• Infrastructure: training and inference including model versioning with an immutable audit trail and a real-time inference targets
• Regulatory inputs: technical contributions to the regulatory pathway, including model cards and design history file documentation, What was the root cause, how did you diagnose the issue, and what changes ultimately improved performance?
• 3. Describe your experience training and validating image or video models.

Include the frameworks and tools you have used (e.g., PyTorch, TensorFlow, OpenCV), your approach to preventing data leakage, handling class imbalance, and validating model performance on unseen data.

• Have you deployed deep learning models into production?

If yes, briefly describe:

The application Real-time or latency requirements Your role in deployment How you monitored and versioned models after release

• Based on the role description, why are you interested in this opportunity, and what unique experience do you believe prepares you to succeed in a founding technical leadership role?

Do you have experience in Test validation method?, * Proven track record training and validating convolutional neural networks for image or video or related tasks, with named projects and measurable results you can speak to in detail

• Strong command of computer vision and deep learning fundamentals, including transfer learning and modern backbone architectures
• Familiarity with multimodal fusion strategies (early, intermediate, late) and their tradeoffs
• Sound grasp of asymmetric error costs and how they shape loss functions, thresholds, and validation design - with the understanding that domain experts will tell you which errors matter most
• Hands-on experience with class imbalance techniques (weighted loss, focal loss, oversampling) and validation
• Production experience deploying models under real-time inference constraints
• Experience with explainability frameworks (GradCAM or equivalent)
• The judgment to assess technical vendor work critically: to separate credible claims from optimistic ones and validate delivered work rather than accept it on faith

Preferred

• Experience with medical imaging or video-based vision tasks
• Familiarity with FDA SaMD documentation or other regulated software development
• Experience building and leading a small ML engineering team
• MLOps maturity: experiment tracking, model registries, reproducible training pipelines

The model you build is the company. Owning it internally and being able to defend every design decision under regulatory and investor scrutiny is the core of what Vanquish AI is. If you want a role where your architecture decisions directly determine whether a product reaches patients, this is it.

Pay: $135,000.00 - $185,000.00 per year

Application Question(s):

• Tell us about the most complex computer vision or deep learning model you have personally owned.

Please include:

The problem you were solving The size and type of the dataset The model architecture Your specific contributions (versus those of your team) The measurable results Whether the model was deployed into production

• Describe a model that did not perform as expected.

We are a clinical AI company building software that detects cancer in real time during medical imaging procedures. Our flagship product applies deep learning to medical imaging to identify and differentiate solid tumors, cysts, and normal tissue at the point of care. The model fuses multiple modalities into a single diagnostic output. We work alongside leading health systems and academic partners and are building toward FDA clearance as a Software as a Medical Device (SaMD).