CTO / Quant Engineer

Enzian Labs; Zürich; Vor Ort; CHF 180.000 - 240.000; Vollzeit

Zusammenfassung

Enzian Labs in Zürich is seeking a data-driven ML engineer with a strong quantitative background to join a project building AI for private equity. You will own end-to-end ML pipelines, move models from notebooks to production, and work with noisy real-world data.

You will fine-tune foundational models, apply Bayesian methods, and optimize with LoRA, quantization, and efficient training. Proficiency with JAX/NumPyro/PyMC and tools like HuggingFace and Unsloth is required.

Qualifikationen

• Quantitative background at MSc/PhD level in computational finance, statistics, applied math, physics, or ML.
• Experience with probabilistic programming and Bayesian inference (JAX, NumPyro, PyMC).
• Hands-on experience building production ML pipelines for real-world data.
• Fine-tuning large models (LLMs/vision) for domain-specific tasks.
• Strong familiarity with the modern AI stack (HuggingFace, Unsloth, PyTorch).

Aufgaben

• Own end-to-end ML lifecycle from model prototyping to production deployment.
• Architect data ingestion pipelines that handle noisy inputs (scanned PDFs, inconsistent reports, missing data).
• Fine-tune foundational models (LLMs and vision) for domain-specific tasks.
• Apply efficiency techniques (LoRA, quantization) to make large-scale AI cost-effective.
• Employ Bayesian inference to model uncertainty in private market valuations.

Kenntnisse

Probabilistic programming

Bayesian inference

Production ML pipelines

LLMs fine-tuning

HuggingFace ecosystem

Unsloth tooling

PyTorch familiarity

Ausbildung

MSc/PhD in computational finance, statistics, applied math, physics, or ML

Tools

JAX

NumPyro

PyMC

HuggingFace

Unsloth

PyTorch

Jobbeschreibung

The Project

We are building an AI system that is the backbone for the private equity industry.

The Role

Quantitative background (MSc/PhD in computational finance, statistics, applied math, physics, or ML). Fluent in probabilistic programming (JAX, NumPyro, PyMC). Hands-on experience building data pipelines on real-world messy inputs. Has fine-tuned large models for domain-specific tasks using frameworks like Unsloth or HuggingFace. Thinks in distributions, not point estimates. Uncomfortable when a system returns a number without a credible interval.

What You Will Do

• Build Production Pipelines: Take ownership of the end-to-end ML lifecycle. You will transition models from local Jupyter notebooks to scalable, production-ready systems.

• Tame Messy Data: Architect data ingestion pipelines capable of handling noisy, real-world inputs—including scanned PDFs, inconsistent reporting formats, and missing data points.

• Leverage Foundational Models: Fine-tune LLMs and vision models for domain-specific financial tasks.

• Optimize for Efficiency: Apply techniques like LoRA, quantization, and efficient training loops using frameworks like Unsloth and HuggingFace to make large-scale AI practical and cost-effective.

• Apply Advanced Mathematics: Utilize Bayesian inference and probabilistic programming to model uncertainty in private market valuations.

What We Are Looking For

Foundations

• Quantitative background (MSc/PhD in computational finance, statistics, applied math, physics, or ML)
• Experience with probabilistic programming and Bayesian inference (JAX, NumPyro, PyMC)

Experience

• Engineering Chops: Proven experience building production pipelines. You know firsthand the critical difference between a proof‑of‑concept demo and a resilient production system.
• Applied AI/GenAI: Hands‑on experience working with foundational models. You have successfully fine‑tuned LLMs.
• Resourceful Tooling: Deep familiarity with the modern AI stack (HuggingFace, Unsloth, PyTorch, etc.) and a knack for maximizing model performance on a startup budget.

Bonus Points

• Comfortable with agent‑based modeling and economic simulation
• Familiarity with financial concepts (NAV, IRR, fund structures) — PE experience a plus but not required