Projects
Systems I've Built & Architected
Production systems powering Gainsight's enterprise platform — from data pipelines to AI-powered search.
Across Two Product Lines
My Contributions to Gainsight's CustomerOS
AI · SEARCH · LLM
Community Search & AI Platform
CC (Customer Communities) · Three AI features, federated search, and a provider-agnostic AI orchestration layer
The AI brain of Gainsight Communities. A provider-agnostic AI orchestration layer powers three flagship features through a factory pattern — primary + fallback LLM chains with automatic cross-provider retry, bidirectional payload transformation for provider compatibility, and graceful degradation on any failure. Synchronized Search Service indexes 25M+ records from 5 federated sources to Algolia with zero-downtime index swaps.
PHP (Yii/Symfony)Algolia AskAIOpenAIGeminiLlama 3 70BAWS BedrockSSE StreamingPythonRedisS3 ParquetAthena
- Cross-provider LLM fallback — dedicated adapter per provider handles payload asymmetry; automatic retry across providers even on client errors to prevent single-provider outages
- SSE streaming with thinking-phase deduplication — stateful stream processor filters duplicate reasoning blocks per model family while preserving final output fidelity
- Domain-driven prompt architecture — composable prompt traits (model, version, temperature, reasoning effort) enable per-feature model versioning and A/B testing without code duplication
- Dynamic rate limiting with zero-restart retuning — atomic Redis counters (no locks, no races), thresholds configurable at runtime so ops can retune without a deploy
- Zero-downtime full sync — bulk-index 25M+ records into temp Algolia index, atomic rename over production. No search outage during reindex
- Analytics data lake on S3 Parquet + Athena with tag expression consolidation reducing Algolia API calls by 90%+
MCP PROTOCOL · PYTHON · AI INFRASTRUCTURE
DCH Search MCP Server
CC · Production Model Context Protocol server — 21 tools making community knowledge accessible to AI assistants
One of the early production MCP deployments integrating Claude, ChatGPT, and Cursor with a real enterprise search backend. 21 read-only tools expose community discussions, knowledge base articles, Skilljar courses, and search analytics — all through permission-scoped Algolia tokens that enforce the same access rules as the frontend UI.
Python 3.11FastMCPSSE TransportJWTRedisAlgoliaECS FargateasynciostructlogPydantic
- 21 read-only tools across 4 backends — unified search, community REST API, education platform, and search analytics — behind a single MCP interface
- Proactive token refresh before expiry — request-time never pays fetch latency. Role resolved from the search service at token-time, not from cached claims (survives long-lived token lifespans)
- Permissioned search at the token layer — access control baked into scoped search tokens, not post-filtered. MCP and frontend receive identical security semantics
- Deliberate availability vs safety tradeoffs per endpoint — revocation checks degrade gracefully while role-based checks remain strict
- Connection-scoped context propagated to all child tool calls — auth context and tokens travel the entire request lifecycle without threading issues
DISTRIBUTED SYSTEMS · DAG ENGINE
Salesforce
HubSpot
Snowflake
BigQuery
Databricks
SAP
Zendesk
Zoom
MS Teams
Intercom
MySQL
PostgreSQL
MSSQL
GA4
Data Highway — DAG-Based Pipeline Engine
CS (Customer Success) · The compute backbone processing 61B+ records/day
A distributed data pipeline platform for enterprise SaaS. Ingests data from tenant systems via 35+ connectors, runs transformations through a SQL engine, and loads results into analytics data stores and warehouses. A directed-acyclic graph of tasks coordinated by leader election and workflow orchestration across a horizontally-scalable worker cluster.
Java 8Vert.xApache DrillZookeeperPython/CeleryPostgreSQLRedisAWS SQSS3
- Processes 61B+ records/day across 600+ enterprise tenants at 98.7% job success rate
- Multi-worker architecture: DAG Workers (Java/Vert.x) + Dynamic Workers (35 isolated connector JARs) + Python/Celery ML Workers + Apache Drill SQL engine
- 60+ custom Apache Drill UDFs for tenant-specific data transformations
- Horizontally-scalable cluster coordinated by Zookeeper leader election and distributed workflow orchestration
- Failure-injection tasks (OOM, disk exhaustion, FD leak) as production resilience probes
35+ Enterprise Connectors
SalesforceHubSpotSnowflakeBigQueryDatabricksSAPZendeskZoomMS TeamsIntercomMySQLPostgreSQLMSSQLGA4+ 21 more
SECURITY · GRAPH DB
User Permissions (RBAC)
CS · Graph-native access control powering authorization across 7+ services
Centralized multi-tenant RBAC service. Permissions stored as a graph — a single traversal answers "can user U do action A on resource R?" in milliseconds, with Redis-cached evaluations and expression-based policy conditions. The permission SDK is consumed by 7+ services across the platform.
Neo4jVert.xRedisMessage QueuePostgreSQL
- Graph-native model: resources, policies, groups, and manager chains as traversable edges — single graph traversal resolves access in milliseconds
- Targeted Redis invalidation — editing a sharing group invalidates only affected members × resources, not the full cache
- Row-level data enforcement — restricted data never leaves the database, access control enforced at SQL query-build time
- Permission SDK consumed by 7+ platform services: analytics, query federation, reporting, API gateway, data store, catalog, and cross-product bridge
- Heartbeat-based leader election for worker coordination with automatic cleanup of orphaned graph nodes
ANALYTICS · WAREHOUSING
Usage Data Model (UDM)
CS · Multi-tenant analytics orchestration across dual warehouses
Full-lifecycle analytics engine — onboarding, schema definition, ingestion from 10+ sources, transformation, and warehouse storage. Powers Gainsight dashboards with per-tenant dynamic DDL and dual-warehouse abstraction (PostgreSQL for small tenants, Redshift for high-volume).
Vert.xRedshiftPostgreSQLMessage QueueRedis
- ~25 pluggable task builders — adding a new data source is a single builder class
- Dual-warehouse abstraction — PostgreSQL for small tenants, Redshift for high-volume OLAP, switched transparently
- Permissions injected at query-build time — restricted data never leaves the database
- Cross-system DAG submission to the pipeline engine with retry and backoff
- Secure multi-tenant data exchange with per-tenant credential isolation
PUB/SUB · EVENT-DRIVEN
Event Processing
CS · Four-tier async pub/sub with durable delivery and cross-region failover
Async event fan-out with durable delivery. Publishers emit once; subscribers register for events with optional delayed delivery and data enrichment. Four-tier architecture with purpose-matched runtimes: Receiver → Router → Relayer → Broadcaster.
Vert.xAWS SQSApache HttpAsyncPostgreSQLRedis
- Cross-region SQS failover — automatic switch to secondary queues on publish errors
- Self-healing delivery: dedicated recovery process detects and recovers orphaned delivery rows after worker crashes
- Blocked-tenant handling isolates noisy tenants from starving other tenants' event delivery
- Enrichment loop — events decorated with business context from the analytics data store before delivery
- Tiered architecture: Receiver → Router → Relayer → Broadcaster with purpose-matched runtimes
COMPLIANCE · OBSERVABILITY
Audit Service
Centralized event capture & compliance trail
Reactive audit service for compliance trails. Fire-and-forget writes via Vert.x event bus decouple caller latency from Postgres persistence. Per-tenant tables created lazily on first write. Used by every service in the platform via a lightweight audit client SDK.
Vert.xPostgreSQLRedis
- Async write via event bus — HTTP 200 returns before DB insert completes
- Random-tenant retention cleanup with +10 day safety buffer
- Separate read/write DB pools to prevent analytics from starving ingestion
INFRASTRUCTURE · CONFIG
Config Server
Centralized configuration for all CS services
Spring Boot + Vert.x hybrid config service. Every platform service pulls runtime configuration from here — DB strings, feature flags, encrypted secrets — with full audit trail and periodic refresh. The config client SDK is embedded across all Java services.
Spring BootVert.xPostgreSQLAES Encryption
- Upsert + append-only version history table in single transaction
- Database-enforced integrity with case-insensitive keys and FK cascades
- Redis-cached auth lookups with fail-fast startup ping
