Key Milestones

MilestoneTimelineDependency
Product v1AugustGoing full-time
Initial cohort sourcedAugust--
Funding--Funding
US market entryQ4 2026US Plan

Hiring Thesis

Hiring Thesis

Tact is trying to turn subjective human judgment into infrastructure AI systems can learn from. That requires a team that can reason from first principles, but also operationalize those principles into data schemas, product surfaces, evals, model improvements, customer proof, and legally usable datasets.

The hiring bar is not intelligence alone. It is the ability to preserve nuance while making it measurable.

We are not looking for people who think taste can be reduced to a single score. We are looking for people who understand that judgment is contextual, disputed, process-driven, and only useful to AI systems when captured with the right traces, outcomes, and evaluation design.

They should be obsessive about what the atomic data unit for Tact is, what makes traces valuable, what labels are noisy, what can be learned, and what will be garbage.

Member of Operations/ Business Staff

Operations Lead

  • I called the shot at Fluency and it's been one of the highest ROI calls - hire a Lawyer to be your ops lead.
    • Their rate of learning is exponential, they can draft, read, review contracts across the entire stack, pick up any operation task (onboarding, payments, anything outside of product and growth) and execute it right almost every time with 0 prompting.
    • Tact's operations surface is legally dense from day one: contributor rights, customer data, de-identification, model training permissions, payout terms, enterprise contracts, and provenance. A lawyer-operator can turn that risk into operating leverage.
Members of Technical Staff

ML Engineers, AI Engineers, Data Engineers, Product Engineers (Fullstack)

Every technical hire must be able to improve one part of the loop: interface → behavior → trace → dataset → eval → model/product improvement.

e.g.

  • Quant traders, statisticians, people who may not come from traditional backgrounds but can approach the novelty of the taste problem from first principles, this can be the long term unlock for innovation and solving the toughest problems we will face. We cannot all be solving the taste and judgement problem from the same set of beliefs, actions and thesis. The more unique backgrounds and systems of thinking we can bring into the room, the better.
  • Members of Technical Staff from frontier labs looking for early stage opportunities, PhD and Masters level AI researchers
RoleMust own
ML / AI engineerTurning traces into classifiers, judges, reward signals, rerankers, evals, and training data
Data engineerProvenance, versioning, schemas, de-identification, QA, packaging, dataset lineage
Product engineerThe creative/work interface that captures real decisions rather than artificial labels
Research-minded engineerExperiments that show whether judgment traces outperform preference labels or generic prompting
Finding a Co-founder

The most important hire is a technical cofounder who can become the owner of Tact's data/model architecture. They should be stronger than me in implementation, fluent in modern post-training and evals, and capable of turning subjective judgment into measurable systems without flattening it into a fake universal score.

They should have founder energy, product taste, and enough intellectual independence to understand where RLHF, preference aggregation, and generic taste philosophies break down. Ideally, they have worked near frontier labs, post-training, evals, fine-tuning, or applied AI products, and can help recruit exceptional technical talent.

Applied ML + evals + data infrastructure + product taste + founder speed.

The first talent search should be heavily US-weighted, especially around SF and frontier-lab networks.