Data Scientist/Machine Learning Engineer

Sumble Inc; United States; On-site; USD 70,000 - 90,000; Full time

Benefits offered by this job

Medical, dental, and vision benefits

401k plan

Target 4 weeks PTO

Job summary

A data-focused technology company is seeking an individual to finetune small language models and enhance existing data quality. The role involves data scrubbing, normalization, and pushing solutions into production environments. Ideal candidates should have Python experience and be familiar with cloud platforms and databases. The company offers competitive benefits, including medical, dental, and vision coverage, and 401k.

Qualifications

• Experience in finetuning machine learning models.
• Ability to improve data quality through scalable methods.
• Proficient in Python and familiar with essential frameworks.

Responsibilities

• Finetune small language models.
• Improve the quality of existing data using scalable approaches.
• Add new signals and associate them with existing ontology.

Skills

Finetuning small language models

Data scrubbing and normalization

Data pipeline management

Python

Familiarity with cloud platforms (GCP)

Working with databases (PostgreSQL, DuckDB)

Experience with frontend frameworks (React, Typescript)

Tools

PyTorch

Huggingface

FastAPI

Job description

Sumble is building a knowledge graph from web data with a first focus on data for go-to-market teams. We use sources like job posts and resume data to identify things like org structure, tech stack, and key projects (e.g., GenAI initiatives, cloud migrations). Our product already has strong product-market fit, early revenue, and happy customers — and now we’re ready to accelerate.

Our long-term vision is to become the primary destination for accessing high-quality web data. Try the product at sumble.com.

Our Team: We are a team of 15, including 10 engineers with experience at companies such as Google, Meta, Stack Overflow, and Kaggle.

What you’ll do

• Finetuning small language models
• Improving the quality of existing data using scalable approaches. Examples include: making sure URLs are associated the right company, we have the correct HQ address, we have mapped parents-subsidiary using techniques like LLM validation, SERP, and triangulating across sources.
• Adding new signals: this usually involves scrubbing, matching and normalizing new signals and matching to our existing ontology
• Pushing solutions into production environments, which may involve touching data pipelines and/or backend systems

• Located within Americas timezones

More about Sumble

Our Tech Stack:

• ML/Data: PyTorch, Huggingface, Gemma models, LORA, VLLM, Skypilot, Marimo
• Languages & Frameworks: Python, FastAPI, React, Typescript
• Cloud Platform: Google Cloud Platform (GCP)
• Databases: PostgreSQL, DuckDB
• Infrastructure: Cloud Run
• Product/Design: Figma, Vercel V0

Challenges We Tackle:

• Transforming noisy datasets into high-quality data products
• Running expensive analytics computations efficiently
• Managing the complexity of a growing number of data sources, machine learning models, and large data operations
• Create a great PLG experience with upsell pathways

• Medical, dental, and vision (US)
• 401k (US)
• Target 4 weeks PTO