Skip to content
Search jobs inside your AI agent. Claude, ChatGPT, Claude Code, Codex, Gemini, Cursor and Grok
Connect

Mappa

NewBe an early applicant

Ingeniero de aprendizaje automático

Posted

Description

Senior Applied AI Engineer

Voice & Behavioral Matching

Location: Remote

Compensation: $4,000–$6,000


About the role

We build AI systems that use voice and assessment data to help connect talent with roles and teams. We already have a working voice model in production whose outputs feed our matching layer.

We are looking for a Senior Applied AI Engineer to own the evolution of this system: extracting signals from proprietary instruments, interpreting results using our theoretical framework, and developing matching across behavioral requirements, team dynamics, and cultural alignment.

You will work across data, predictive models, LLM workflows, evaluation, and production in a fast-moving environment with low bureaucracy. Your responsibility is to identify what limits the system, choose the right technical approach, ship improvements, and demonstrate their impact.


What you will do
  • Design the architecture connecting signal extraction, behavioral interpretation, and matching. Decide where to use deterministic logic, predictive models, pretrained models, or LLMs based on the problem and measured results.
  • Build LLM workflows that interpret assessment results using our proprietary framework. Structure the relevant knowledge and context, produce traceable outputs, and handle conflicting or insufficient evidence.
  • Develop candidate-to-role and candidate-to-team matching across behavioral requirements, work practices, and team context. Make each layer explainable, testable, and measurable as outcome data becomes available.
  • Own data pipelines for collection, validation, labeling, versioning, and dataset creation. Improve the quality of training and evaluation data, including scarce, delayed, or noisy labels.
  • Train, evaluate, and maintain predictive models. Diagnose failures across audio processing, transcription, diarization, feature extraction, model outputs, LLM interpretation, and downstream matching.
  • Build reproducible evaluations for each component and the full system. Benchmark proposed changes against production, measuring predictive quality, interpretation quality, latency, reliability, and failure modes.
  • Work closely with our psychometrician to translate measurement requirements and proprietary theory into model targets, interpretation specifications, and validation criteria.
  • Deploy and monitor improvements, optimize inference cost and speed, and simplify infrastructure. Keep data, models, prompts, and knowledge versions traceable so the system is easier to debug and evolve.


What we are looking for
  • Experience building and maintaining LLM applications in production, including context design, retrieval, structured outputs, and evaluation.
  • Strong production ML experience: you have trained, deployed, and maintained predictive or recommendation models and can explain how you measured their performance.
  • Strong Python and SQL, with practical experience in at least one modern ML framework such as PyTorch, TensorFlow, or JAX.
  • Experience with data modeling and ML data pipelines, including dataset quality, labeling, versioning, and working with incomplete or noisy data.
  • Hands-on experience with audio or speech systems and the ability to investigate failures across preprocessing, transcription, features, and model predictions.
  • Strong experimental design and model evaluation skills. You can question the label, identify data leakage, investigate cohort differences, and distinguish a measured improvement from a misleading metric.
  • Ability to make architectural decisions across data, model development, LLM workflows, serving, and operations, and communicate technical tradeoffs clearly.


Nice to have
  • Deep experience in speech modeling, acoustic representations, diarization, or audio feature extraction.
  • Experience with psychometrics, behavioral assessment, talent matching, or recommendation systems evaluated against real outcomes.
  • Fine-tuning, multimodal systems, model routing, or specialized models for different cohorts.
  • Quantization, distillation, or smaller architectures, with evidence of reduced cost or latency while maintaining acceptable predictive quality.
  • AWS, Modal, containerization, asynchronous jobs, model serving, or production monitoring.


How you work

You take ownership from problem definition through production. You can turn an ambiguous request from founders, Product, or Ops into a small executable experiment, make progress with incomplete data, and explain your assumptions and tradeoffs.

You use AI tools actively for prototyping, data exploration, coding, and debugging, and verify their outputs. You can describe what you personally built, what broke, and how you determined whether a change improved the system.


Your first 90 days

Within your first quarter, we expect you to:

  • Establish a reproducible baseline and identify the main limitations across signal extraction, interpretation, matching, and production.
  • Agree on evaluation priorities with our psychometrician and turn at least one ambiguous request into an executable experiment.
  • Ship at least one measurable improvement to the voice model, interpretation layer, matching, or pipeline, with before-and-after evidence and clear tradeoffs.
  • Work across data, model development, and inference, leaving the system easier to operate or debug and a clear plan for the next experiments.
  • Demonstrate practical, verified use of AI tools in your development workflow.



Details
  • Full time (monday- friday)
  • Remote
  • Payment in USD [4500-6000 USD]


Skills

Apply

See also