AI Engineering at BuildBetter.ai

As an AI Engineer at BuildBetter, I worked on integrations, sync infrastructure, CI/CD tooling, and internal developer workflows. BuildBetter is a customer intelligence platform that ingests company data (calls, recordings, interviews, feedback) and turns it into actionable product-ops signals.

Impact

• 844+ commits and 60+ PRs to the main repository
• Expanded data connectivity by building 4 third-party integrations, enabling customers to import data from additional platforms
• Reduced manual work for data imports with AI-powered field classification that automatically maps columns
• Improved engineering velocity through Claude Code skills and slash commands used daily by the team
• Increased CI reliability by enabling previously-skipped E2E tests and adding automated code review checks
• Established integration standards with documentation and checklists that streamlined future integration development

Areas of Contribution

Third-Party Integrations

I built several integrations following OAuth2 and incremental sync patterns, enabling customers to automatically import and analyze data from their existing tools.

Circle.so Integration

Built a complete integration for community platform data.

Challenges solved:

• Worked around API limitations that didn’t support date filtering by implementing client-side early termination optimization
• Implemented per-space incremental sync with database-backed cursor tracking (unlike single-cursor integrations)
• Added cancellation checkpoints for graceful shutdown without losing progress
• Implemented flexible backfill options (none, time windows, or full history)

Notion Integration

Built OAuth2 integration for importing Notion workspace pages as searchable documents.

Features:

• OAuth2 flow with automatic token refresh
• Incremental sync that skips unchanged pages
• Title extraction handling various Notion page structures
• Duplicate detection for workspace re-authentication
• Settings UI for workspace management and import toggles

Front Integration

Built the Front (customer communication platform) integration from scratch.

Features:

• Full OAuth2 implementation with token refresh
• Conversation sync with message threading
• Correct author attribution for inbound vs outbound messages
• Incremental sync using timestamp-based filtering

Attio Integration

Built CRM sync for importing contacts and companies.

Features:

• Sync service with pagination
• Error handling with comprehensive test coverage
• Author resolution with fallback chain

Integration Fixes

• Gong: Fixed category filter that incorrectly rejected calls without optional metadata fields
• HubSpot & Zendesk: Fixed ticket sync to handle tickets without authors gracefully

AI Features

Custom AI Agents (Draft Iteration)

Built an early draft iteration of the Custom AI Agents feature, which enables AI-powered conversations for analyzing customer signals. This exploratory work helped inform the direction of the feature before it was rebuilt by another engineer.

AI-Powered Field Classification

Built an LLM-powered system for automatically classifying and mapping fields during data imports.

Features:

• Analyzes column names and sample data to determine field types
• Matches against existing organization fields for consistent naming
• Returns confidence scores for each classification
• Fallback UX for edge cases where AI classification is uncertain
• Auto-accept for high-confidence mappings

UI/UX work:

• Column name truncation for long headers
• Sample row preview
• Persisted user mappings through async processing

Developer Tooling

Created tools and documentation to standardize how the engineering team builds features and debugs issues.

Claude Code Skills

Built reusable AI-powered development tools:

• Development environment startup
• CloudWatch log querying for staging/production
• Database querying via GraphQL
• Migration generation
• REST endpoint scaffolding
• Permission configuration
• Circular dependency diagnosis
• Admin UI scaffolding
• CI failure analysis

Slash Commands

Created workflow automation commands including:

• Integration planning - Interactive command that researches provider APIs and generates planning documents with authentication strategy, entity mappings, sync strategy, and implementation checklists
• Implementation planning - Research-driven planning workflows
• Feature specifications - Spec documents with user journeys
• Error review - Periodic error analysis and recommendations

Integration Pattern Documentation

Created documentation for building integrations:

• Architecture patterns and tech stack requirements
• Implementation checklists covering infrastructure, migrations, services, workers, and frontend
• Acceptance criteria standards with unit test and telemetry requirements

CI/CD & Testing

Automated Code Review

Analyzed historical PRs to extract actionable heuristics that run automatically in CI:

• Architecture migration rules (ORM migration, API patterns)
• Type safety rules
• Security checks (permissions, authentication)

Benefits:

• Instant feedback without LLM latency for obvious issues
• Consistent enforcement of codebase standards
• Reduced review cycles

E2E Testing Infrastructure

Built infrastructure to run previously-skipped E2E tests in CI:

• Testcontainers pattern for database isolation
• Dynamic test context (no hardcoded IDs)
• Proper external service mocking
• Dedicated E2E configuration and CI job

Additional CI Improvements

• Pre-flight validation before expensive tests
• Timeout and concurrency controls
• Job summaries and annotations
• Draft PR skipping
• Quality gates for architecture patterns

Test Coverage

Added unit tests for integrations including error handling, sync services, and auth decorators.

Infrastructure Work

Queue Reliability

• Built monitoring for failed message detection with appropriate alert thresholds
• Improved visibility management with graceful handling of expired messages
• Structured logging with job context for debugging

Database Migration

• Contributed to ORM migration initiative with integration tests
• Extracted database queries to repositories for testability
• Type-safe SQL expressions

Observability

• Improved telemetry consistency across services
• Added error tracking for previously unmonitored failures
• Structured error logging

Technologies Used

• Languages: TypeScript
• Frameworks: NestJS
• Databases: PostgreSQL, Kysely
• Message Queues: AWS SQS
• Infrastructure: AWS CloudWatch
• Testing: Jest, Testcontainers
• CI/CD: GitHub Actions