Take control of your data

Salt Data Labs helps companies make their data valuable.

Salt Data Labs offers a unique blend of advising, hands on machine learning model development and deployment, statistical analysis, and business problem solving for all industries. We work to understand our clients’ business, needs, goals, and processes to provide accurate, comprehensive, and data driven solutions.

Our Services

SDL offers a wide range of data and research services including development, trainings, large scale strategic planning, machine learning based product development, hiring and sourcing, and everything in between.

Our Services

Blog

Check out what we’re up to or learn about the latest in Machine Learning and Data Science.

Read Our Blog »

The Problem

A healthcare recruiting firm hired Salt Data Labs to develop an entity resolution model for a rapidly growing database of candidates and employers, and overhaul an outdated ETL pipeline.

Our Solution

Developed lightweight, explainable algorithms for deduplication, record linkage, and canonicalization.

Rebuilt the companies ETL pipeline, ported from R to Python, automated to run on a schedule and in response to data driven triggers. Implemented best in class logging, monitoring, and alerting.

The Impact

Project timeline: ~25 person days
Payback period: < 2 weeks

Decreased ETL runtime over 85% (hours to minutes), decreased loading failures by more than 50%, and eliminated the need for manual runs. The company is now saving valuable engineering hours every week, while realizing reduced compute and data storage costs.

Reduced human time spent on entity resolution tasks by over 90%, reduced data storage costs, unlocked the ability to link records across disparate datasets.

The Problem

Salt Data Labs was hired by a marketing automation company to upskill their existing DS team, and figure out if a large language model (LLM) based content automation feature was worth building.

Our Solution

Created tailored trainings for Junior and Senior data scientists and engineers on model deployment, deep learning for LLMs, content generation, active learning, and user research. Held weekly office hours for internal Data Science team and stakeholders.

Worked with internal Data Science team to analyze SMS, email, and long form marketing copy, define content quality metrics, experiment with multiple LLMs, collect and analyze user feedback, and ultimately design and build a GPT-based content generation engine.

The Impact

Our client was able to avoid a costly hiring process, and transition their most senior Data Scientist to work on this cutting edge project.

As they figure out their product roadmap, the company has clear evidence from robust user research and testing that their efforts must include investment in generative AI development.

The Problem

Salt Data Labs was hired by an online remittance company to analyze 6 years of Net Promoter Score (NPS), customer survey, and retention data. The data comprised 135 countries, multiple languages, and over 1 million customers.

Our Solution

Created explainable and interpretable NLP models for topic extraction and sentiment classification to surface trends, pain points, and highlights from NPS survey data. Created derived features from text data to support richer analysis of customer behavior and outcomes.

Designed and implemented a suite of statistical tests to qualify the effect of pricing and fee changes on customer satisfaction, complaints, usage, and NPS score.

Contributed to presentations for diverse audiences to showcase results, examples, and product recommendations.

The Impact

Project timeline: 7 person days
Payback period: ~1 week

Our client gained a deep understanding of how pricing and fee changes impact customer retention, sentiment, satisfaction, complaint areas, and key metrics.

They used our derived features for segmentation, experimentation, and consumer insights reporting.

The company now has a reusable analysis pipeline that is used to measure the effect of various product changes, saving hours of manual work for each experiment while increasing insights significantly.