Engineering & Technology - Engineering
Multimodal AI Engineer
IC / Practitioner
Independently auditedBuilds and operates the software the business runs on. Owns a defined area of the function and the outcomes it produces.
Audited package
Built, independently audited role package. Everything on this page is written for this seat.
Your Stack (Swappable)
This role works against whatever tools you already run. We list the tool categories the role operates in, never a named vendor, so you are never asked to replace a system that works.
- Learning & Development (LMS)
- Vision and image processing
- Audio / Video Production
- Video processing
- AI / LLM Platform
- Experimentation & Feature Flags
- Serving hand-off
Deliverables
- Design documents
- Review checklists
- Technical standards
- Delivery and sequencing plans
Responsibilities
- Design multimodal architectures and experiments across modality-specific encoders, projectors, fusion, attention, token budgets, and output heads.
- Build versioned preprocessing and modality contracts for text, image, audio, video, temporal sampling, resolution, normalization, and missing-input behavior.
- Curate and validate paired, interleaved, and instruction data with balance, provenance, consent, quality, safety, and leakage controls.
- Fine-tune and evaluate models with controlled baselines, ablations, loss weighting, curriculum, freeze or unfreeze choices, and reproducible runs.
- Measure grounding, cross-modal consistency, conflict resolution, modality use, hallucination, calibration, safety, and robustness under corrupt or missing inputs.
- Diagnose representation collapse, modality dominance, preprocessing mismatch, temporal misalignment, dataset shortcuts, and evaluation blind spots.
- Prepare a serving contract covering media limits, preprocessing, batching constraints, model artifacts, memory estimates, latency targets, fallback, and quality guardrails.
- Partner with Research, Data, Evaluation, Safety, Privacy, Product, Applied Inference, and ML Platform teams without taking their decisions.
The essence of this seat
The Multimodal AI Engineer is individual contributor who is the engineer who builds and evaluates models that combine text with image, audio, video, or other modalities, owning the multimodal data, alignment, fusion, fine-tuning, and model-behavior engineering path.
Signature move
Write the modality contract first: accepted inputs, preprocessing, temporal or spatial sampling, missing modalities, output schema, and evidence the answer must ground in
What good looks like
treats preprocessing as model behavior, versions temporal sampling and tokens, and requires executable parity across training, evaluation, and serving
Decision rights
- - the technical design and implementation choices inside the bounded modality-specific encoders and preprocessing, representation alignment, fusion and projector architectures, paired and interleaved datasets, multimodal instruction tuning, grounding, cross-modal consistency, modality balance, safety, evaluation, ablation, and model hand-off for production serving work assigned to this seat
- - the evidence, tests, safeguards, and operational artifacts needed to demonstrate this outcome: a multimodal model uses each modality for the intended evidence, aligns and fuses representations coherently, performs across realistic combinations and missing-input cases, and ships with reproducible behavior, limitations, safety evidence, and a clear serving contract
- - when a change inside the delegated scope is ready for review, canary, rollback, or further investigation
What this role does
The Multimodal AI Engineer is individual contributor who is the engineer who builds and evaluates models that combine text with image, audio, video, or other modalities, owning the multimodal data, alignment, fusion, fine-tuning, and model-behavior engineering path. The role is accountable for the engineering work required so a multimodal model uses each modality for the intended evidence, aligns and fuses representations coherently, performs across realistic combinations and missing-input cases, and ships with reproducible behavior, limitations, safety evidence, and a clear serving contract. Its working domain includes modality-specific encoders and preprocessing, representation alignment, fusion and projector architectures, paired and interleaved datasets, multimodal instruction tuning, grounding, cross-modal consistency, modality balance, safety, evaluation, ablation, and model hand-off for production serving. A strong practitioner turns that domain into explicit contracts, measurable evidence, controlled changes, and clean hand-offs. The role reasons from the actual artifact and failure boundary rather than from a fashionable tool. It makes assumptions visible, tests the load-bearing ones, and distinguishes what is Verified, Inferred, and Not-yet-verified. The role remains at IC altitude: it owns technical execution and bounded decisions, advises leaders and neighboring specialists, and escalates product, people, budget, policy, legal, risk, and cross-organizational calls. The nearest boundaries are deliberately strict. Distinct from AI/ML Engineer: this seat owns the specialized multimodal model, data, alignment, and cross-modal evaluation problem; that seat owns broader applied-ML product delivery across model types. Distinct from Machine Learning Engineer: this seat owns multimodal architecture and behavior-specific experimentation; that seat owns general training infrastructure, model lifecycle, and production ML systems outside this specialization. Distinct from Software Engineer - Applied Inference: this seat builds, fine-tunes, and validates the multimodal model and its input-output contract; that seat optimizes runtime batching, memory, kernels, quantization, throughput, latency, and serving reliability. Those titles still receive their own complete packages under the 1:1 policy.
Best-fit customer
engineering teams that want consistent judgement applied to design, review, and delivery
Need more than one role?
Buying the whole team or the complete department lowers your effective cost per role at every tier of the ladder.