Founder-led. Remote-first. Anti-hype.
Anchor Data Labs is a data consulting and contracting firm run by Pranav Shah. We do the unglamorous, high-stakes work underneath AI: governing data, migrating it without breaking reporting, automating what's manual, and re-engineering the content an AI agent retrieves from.
We work both ends of the market, from large enterprise engagements to small outsourced projects, remotely for clients anywhere. When a piece of that work becomes repeatable, we turn it into a self-service tool. ContextScore is the first.
Every engagement runs the same way. We diagnose before we prescribe, scope to what you actually need, deliver, and hand it over in a state your team can run without us. If we're not the right firm for the job, we say so early.
These are Pranav's prior enterprise and research roles, not engagements delivered by the firm. Anchor Data Labs is new, and applies exactly this experience.
Capital One (US & UK)
Across US analytics, then UK data management. In the US, senior and principal data-analyst work: Databricks and SQL pipelines across S3, Snowflake and data lakes, automation that removed about 50 hours a month of manual prep, and Tableau and Salesforce dashboards for Small Business Banking teams. In the UK, led data-centre exit migrations and, as lead data consultant on the Data Management team, set the data-lifecycle standards: naming and modelling, data-movement and infrastructure choices, and how data is stored and delivered. Documented the exceptions and anti-patterns, and built an internal data-governance agent, packaged as a reusable Claude Skill, that brought that guidance into teams’ own work.
Rio Tinto
Data scientist on industrial time-series: a machine-learning model for operational analysis, computer-vision models for movement, human safety, and object detection, and a Power BI and SQL Server reporting rebuild that replaced spreadsheet planning.
University of Utah
MS Computer Science. As a graduate research assistant for the School of Computing, maintained the school’s website and built visualization tools that let professors track student coursework and progress. Academic grounding in coreference resolution and semantic-similarity methods.
Diagnose before we prescribe
Every engagement is scoped to the problem in front of us.
Evidence over assertion
Every judgment cites its source. When we're unsure, we say so.
Governed by design
Access fits each role. Compliance, privacy, and data-risk reviews are built in where the work needs them.
Models are reviewed, not trusted
Model risk assessment and review before anything ships.
Cost is a design constraint
Spend caps and rate limits from the start, so AI stays predictable.
Standardized by default
Consistent naming, modeling, and data quality, so the work holds up after we leave.