Weld

Weld provides a drag-and-drop interface that abstracts SQL and code-based ETL logic into visual node-based workflows. Users connect source connectors to transformation nodes to destination connectors without writing code, with the platform translating visual configurations into executable data pipelines that run on a managed cloud infrastructure. The builder uses a directed acyclic graph (DAG) model where each node represents a discrete operation (extract, transform, load) and edges define data flow dependencies.

Weld maintains a curated library of 50+ pre-configured connectors for popular business tools (Salesforce, HubSpot, Stripe, Google Analytics, Shopify, etc.) that handle authentication, pagination, rate limiting, and API schema mapping automatically. Each connector encapsulates the source system's API contract, exposing normalized field schemas and available operations (read, write, upsert) without requiring users to understand the underlying API. Connectors use OAuth 2.0 for user-facing SaaS tools and API key/token management for backend systems.

Weld supports both incremental (delta) and full-refresh synchronization strategies, allowing users to configure pipelines that either pull only changed records since the last run or re-sync the entire dataset. The platform uses timestamp-based or cursor-based change detection to identify new/modified records in source systems, reducing data transfer volume and API costs. Schedules are defined via cron expressions or simple UI selectors (hourly, daily, weekly) and executed on Weld's managed infrastructure with automatic retry logic and exponential backoff for transient failures.

Weld provides a visual field mapper that allows users to drag source fields to destination fields, with automatic data type conversion (string to number, date parsing, null handling). The mapper supports one-to-one field mapping, field renaming, and basic transformations like concatenation, substring extraction, and conditional logic via simple UI controls. Under the hood, Weld translates these mappings into transformation expressions that run during the data pipeline execution, converting source data to match the destination schema without requiring SQL or code.

Weld captures detailed execution logs for each pipeline run, including record counts (processed, inserted, updated, failed), error messages, and data quality issues (null values, type mismatches, constraint violations). Users can configure alerting rules (email, Slack) for pipeline failures or data anomalies (e.g., 0 records synced when expecting 1000+). The platform provides a dashboard showing pipeline health, last run status, and historical execution trends, enabling non-technical users to monitor data quality without SQL queries or log aggregation tools.

For systems not covered by pre-built connectors, Weld allows users to define custom REST API connectors by specifying endpoint URLs, authentication method (API key, OAuth, basic auth), request/response schemas, and pagination logic. The platform handles HTTP request execution, response parsing, and error handling, exposing the custom connector as a reusable source or destination in pipelines. This enables integration with niche or proprietary APIs without requiring custom code, though it requires users to understand API documentation and HTTP concepts.

Weld supports upsert (update or insert) operations that prevent duplicate records when syncing data multiple times. Users define a primary key or unique identifier field(s) that Weld uses to detect existing records in the destination system; if a record with the same key exists, it updates the existing record instead of inserting a duplicate. This enables idempotent syncs where re-running a pipeline produces the same result regardless of how many times it executes, critical for reliable data integration without manual deduplication.

Weld allows a single source to feed data to multiple destinations in parallel, enabling one-to-many data distribution patterns. A pipeline can extract data from Salesforce and simultaneously write to a data warehouse, a marketing automation platform, and a business intelligence tool, with each destination receiving the same transformed data. The platform executes destination writes in parallel (where possible) to minimize total pipeline runtime, though failures in one destination don't block others (configurable per pipeline).

Weld offers a free tier supporting a limited number of records per month (typically 10K-100K) and a small number of active pipelines, enabling teams to test workflows before committing budget. Paid tiers scale based on data volume (records synced per month) and number of pipelines, with transparent pricing displayed in the UI. This removes friction for initial adoption but creates cost concerns at scale; high-volume syncs (millions of records) can become expensive compared to enterprise ETL tools with flat-rate licensing.