Senior ML​/AI Engineer

About Poseidon

Poseidon is an a16z-backed startup building a platform that coordinates supply and demand for specialized AI training data. We work with Fortune 500 enterprises and leading AI labs to build and operationalize large-scale, rights-cleared multi-modal datasets and the models that learn from them.

We are seeking a Senior ML/AI Engineer to lead the work of taking cutting‑edge ML (voice and beyond) from prototype → reliable systems → customer‑facing product. This is a senior, highly hands‑on role focused on building production‑quality model and data systems, owning technical direction for key components, and raising the bar on engineering rigor.

A small portion of time can be spent on applied research (e.g., new evaluation methods, fine‑tuning recipes, or model quality studies), but the core mandate is to ship
.

• Own end‑to‑end delivery of ML capabilities into product: define the technical plan, implement, product ionize, and operate systems with clear quality, latency, and cost targets.

• Build and scale evaluation for voice AI and other modalities:

  • Design offline + online evaluation frameworks
  • Create workflows for quality measurement and continuous improvement.
  • Partner with product to translate metrics into product requirements and SLAs.

• Lead fine‑tuning and adaptation work
  :

  • Build and maintain pipelines for supervised fine‑tuning and domain adaptation.
  • Own dataset curation, training data strategy, and reproducibility.

• Engineer data and labeling systems that power learning loops:

  • Design schemas/manifests across modalities and automate validation.
  • Build data quality checks: PII detection, deduplication, drift checks, consensus labeling, gold sets.

• Productionize model and pipeline infrastructure
  :

  • Refactor research prototypes into tested Python libraries, services, and batch jobs.
  • Deploy and operate inference endpoints (real‑time and batch)
  • Optimize for GPU/CPU cost and performance

• Raise engineering standards and mentor
  :

  • Set best practices for testing, CI/CD, code review, documentation, and operational readiness.
  • Mentor other engineers and help unblock cross‑functional execution with researchers, PMs, and ops.

• 6+ years of hands‑on experience shipping ML systems to production (or equivalent depth via impactful projects).

• Expert Python engineering skills, including writing maintainable libraries/services, tests, and performance‑aware code.
• Strong experience with modern deep learning frameworks (PyTorch strongly preferred).
• Proven track record owning production ML systems end‑to‑end, including:
• Data pipelines and training/evaluation workflows
• Deployment (APIs, batch jobs, or streaming inference)
• Observability (metrics, logs, traces), on‑call, and iterative reliability improvements
• Experience with voice AI / speech (ASR, diarization, audio preprocessing, alignment, multi‑speaker challenges).
• Strong understanding of ML evaluation and measurement (dataset design, slice‑based analysis, regressions, and statistical thinking).
• Solid cloud infrastructure experience (AWS, GCP, or Azure), containerization (Docker), and production deployment patterns. Kubernetes experience is a plus.
• Excellent communication: ability to write clear technical plans, make tradeoffs, and align stakeholders.

• Experience with multimodal systems (text + audio + image/video) and building unified data/eval abstractions.
• Experience with distributed training, GPU performance tuning, and large‑scale experimentation.
• Experience with workflow orchestration and distributed compute (Ray, Spark, Dask, Airflow, Flyte, Prefect).
• Familiarity with privacy, security, and compliance concerns in ML systems (PII, rights management, auditability).

• Python, PyTorch, FastAPI
• Docker, Kubernetes, Terraform
• AWS/GCP/Azure managed compute + storage
• ML tooling: MLflow or Weights & Biases, model registries, dataset/versioning tools
• Orchestration:
  Airflow, Flyte, Prefect (or similar)
• Observability:
  Prometheus, Grafana, Open Telemetry, cloud‑native logging

• High leverage: your work will ship into products used by enterprises and leading AI labs.
• Real‑world ML: build systems that connect data → training → evaluation → deployment → feedback loops.
• Ownership: senior engineers here drive architecture and outcomes, not just tickets.

If you’re excited to turn state‑of‑the‑art voice + multimodal ML into reliable products, we’d love to hear from you.