Position Summary

Omnilow AI Ventures is seeking an LLM (Large Language Model) Engineer to help build an LLM that natively understands Omnilow, a novel algorithmic language that is as versatile as natural language. This is an early-stage startup role working directly with the creator of Omnilow, Samuel Sauer. Compensation may include cash, equity, or a combination of both.

We are going to modify an open source LLM so it is able to properly read and write Omnilow. You will be responsible for all technical work involved in building this LLM. Samuel will provide domain expertise on Omnilow, and he (potentially along with others) will be responsible for all non-technical functions.

Responsibilities

• Evaluate and determine which open-weight LLM to use as the initial base model.
• Handle all technical aspects of building this LLM end to end, including environment setup, model modification and training, evaluation, iteration, and preparing a finished, deployable model.
• Document all work clearly and reproducibly so that a technically competent outsider could independently reproduce the results, and so your documentation is suitable for patent and IP protection.
• Modify, fine-tune, and evaluate open-weight LLMs to reliably read and write Omnilow.
• Design and run experiments to measure correctness, consistency, and failure modes.
• Iterate on training approach, data strategy, prompting, and evaluation methodology.
• Investigate representation and tokenization tradeoffs for structured language behavior.
• Contribute to technical roadmap and prioritization in a high-ambiguity startup environment.

Required Qualifications

• Must be fluent in English. The LLM will only be trained on Omnilow using American English.
• Strong technical judgment and independence.
• Clear written communication and comfort working asynchronously.
• Upholds high ethical standards and acts with integrity in all technical and professional decisions.
• Able to pass a background checks. No history of fraud, violence, dishonesty, or other serious ethical misconduct.
• Willing to complete a background check.
• Willing to sign confidentiality and intellectual property assignment contracts.
• Able to receive direct, candid feedback and act on it, while collaborating effectively with others.
• Ability and experience in performing each and every function listed in the Responsibilities section.
• Have your own equipment to perform all functions listed in the Responsibilities section, including a suitable workspace, a very capable PC or laptop, and necessary software. You will not be expected to cover operational costs such as cloud compute.

Preferred Qualifications

• Proven experience taking an open-weight LLM and modifying its weights through fine-tuning and otherwise modifying it to produce a finished, deployable model.
• Ability and experience shipping experiments end-to-end (data, training, eval, analysis).
• Live in the United States.
• Fluent in written American English and is your primary language.
• Experience/Knowledge to take a finished LLM and enable users to pay to use it through an API.
• Experience in Notepad++ especially natural language in Notepad++ using complex parentheticals/brackets for grouping and unambiguous writing.
• Experience using Notepad++, particularly for working with structured natural-language-like text using complex parenthetical or bracket structures.
• Prior early-stage startup experience.
• Strong PyTorch skills and practical transformer experience.
• Interest in language design, structured representations, or reasoning systems.

Time Commitment

• Flexible
• If not joining initially as a partner, the role would begin as a freelancer, contractor, or temporary employee.
• Expected range: 10-60 hours/week initially
• You would largely be self-managed and choose your own hours.

Compensation

Compensation may include cash, equity, or a combination, depending on role scope and mutual fit. The details for compensation are flexible.