About THIA
THIA is transforming how small and medium enterprises build internal applications and automate business processes. Our AI-powered platform enables business experts to create custom applications using natural language, eliminating the need for expensive development teams. We’re well-funded, generating revenue, and solving real problems for companies that need more than off-the-shelf software.
About the Role
This is a hands-on ML role for someone who wants depth in modeling and pragmatism about everything else it takes to ship. You’ll own ML work end to end: framing the problem, picking the right approach (fine-tune, prompt, retrieve, or something cheaper), fine-tuning the language models that power our platform when that’s the answer, building the evals that tell us whether things are getting better, and getting the result into production.
We’re open to a range of profiles. Some people will go deepest on fine-tuning and evaluation; others will be strongest at shipping ML features across the full stack. We’ll shape the scope around where you’re strongest. You’ll work closely with a small, senior team and have direct influence over what we train, how we measure quality, and how we build.
We move fast, keep our codebase clean, and take tech debt seriously.
What You’ll Do
Modeling & Evaluation
- Fine-tune transformer-based models (instruction tuning, LoRA/PEFT, RLHF/DPO, distillation) and ship the result through eval into production
- Make pragmatic tradeoffs across fine-tuning, prompting, and retrieval based on what the problem actually needs
- Design and curate evaluation datasets that meaningfully reflect real customer behavior
- Build and extend evaluation pipelines: offline metrics, regression detection, and LLM-as-judge aligned against human judgment
- Run experiments end to end: hypothesis → controlled comparison → calibrated metrics → decision
Shipping ML Features
- Ship ML features end to end: problem framing, modeling, evaluation, deployment, iteration
- Work with cloud ML infrastructure across training, serving, and monitoring
- Own model-quality monitoring in production and close the loop back into training data
- Drive concrete quality gains through eval results, customer feedback, and prompt iteration
Collaboration
- Work autonomously while staying tightly coordinated with a small, async-first team
- Partner with the platform team on model serving, observability, and the multi-tenant rollout, and with product on what to ship next
- Contribute to architectural decisions and internal documentation
What We’re Looking For
Must-Haves
- Strong Python; working expertise with PyTorch and the HuggingFace transformers ecosystem
- Hands-on experience training or fine-tuning transformer-based models, with at least one result shipped or rigorously evaluated
- Has built or substantially contributed to a model-evaluation pipeline, with experimental rigor: hypothesis design, controlled comparisons, calibrated metrics
- Has carried at least one ML feature through the full lifecycle (data → train → eval → deploy → monitor)
- Comfortable with cloud ML infrastructure (GCP, AWS, or Azure) across training, serving, and monitoring
- Experience collaborating with product or domain experts to ship real features
Nice to Have
- LLM eval and observability work (LLM-as-judge, human-judgment alignment, trace enrichment)
- RAG or retrieval system design experience
- MLOps tooling (model registries, feature stores, experiment tracking)
You Don’t Need
- Mastery of every part of the ML stack. Depth in one area plus working knowledge across the rest is the target.
Our Stack
Python · PyTorch · HuggingFace · Modal · GCP · PostgreSQL / SQLite · Qdrant · Redis · Docker · GitLab CI/CD · Datadog
Compensation & Benefits
The salary range for this position is $110,000 – $160,000 per year. Placement within the range is based on role-relevant experience and skills. This range reflects what we genuinely expect to pay for the position; it is determined by the role and is not negotiable. Compensation also includes early-stage equity, health/dental/vision benefits, paid time off, and a fully remote work arrangement.
Equal Opportunity
THIA is an equal opportunity employer. We evaluate all applicants based on skill and ability to do the work, and we do not discriminate on the basis of race, color, religion, sex, sexual orientation, gender identity, national origin, age, disability, veteran status, or any other characteristic protected by applicable law.
Accommodations
If you need a reasonable accommodation at any point in the application or interview process, let us know and we’ll work with you.
How We Review Applications
We use AI tools to assist in reviewing applications and identifying qualified candidates. All final hiring decisions, including who advances through the interview process and who receives an offer, are made by our team.
A Note on AI
We actively encourage using AI tools to move faster: for coding, experiment setup, analysis, and documentation. Real engineering judgment is still required: to direct AI effectively, catch what it misses, and make the final call before work ships.
What You Gain
- Ownership: end-to-end accountability for model quality and ML features at a growing AI company
- Impact: direct collaboration with leadership and real influence on technical direction
- Growth: a clear path to a lead role as the team expands
- Equity: early-stage equity at an AI startup
- Quality: a clean codebase and a team that takes evaluation rigor and tech debt seriously
