mayurjag98

1. Architected and delivered an async, event-driven job-processing microservice in Go using hexagonal architecture, spanning 30 core service files (adapters, ports, use-cases, tests) and integrating Google Pub/Sub for queue orchestration plus cleanup/status reconciliation; improved modularity and enabled faster, isolated testing across domain logic. 2. Migrated DB connectivity from static credentials to IAM-based authentication with Cloud SQL connector + automatic token refresh, removing hardcoded secret dependency and improving auth resilience for long-running service flows. 3. Standardized release validation and API contracts by adding service-specific Helm values and normalizing route/requestresponse handling for job submission/status flows, reducing config drift and integration friction across consumers. 4. Adopted Cursor as primary AI-assisted IDE for day-to-day development, accelerating feature delivery, refactoring legacy service code, and writing test scaffolding, while using AI-assisted code review and doc generation to speed up PR turnaround and reduce review cycles.

1. Co-developed a retrieval-augmented AI search engine for querying LLM knowledge bases, building embeddings pipelines, vector DB integration (Pinecone and FAISS), and agent orchestration logic end-to-end. 2. Integrated MCP servers into the agent execution flow, enabling tool (resource) discovery and extending agent capabilities without hard coding integrations. 3. Built a model gateway/proxy abstraction layer to decouple agent logic from specific LLM providers, enabling provideragnostic routing, centralized configuration, and observability across model calls. 4. Implemented API-key and service-to-service authentication with credentials managed via secret-management mechanisms, plus OAuth flows for external integrations. 5. Applied standard data protection practices across the stack, TLS in transit, encryption at rest, access controls on vector data stores, and PII masking before data reached external model providers. 6. Built personal AI agents for daily workflow automation (task triage, research summarization) as supporting tooling for the core product.

1. Architected and delivered Switchback Experimentation in Litmus (in-house A/B testing platform), introducing cyclical & randomized models with new data schemas, routing logic, audit tracking, and orchestration in Go & PostgreSQL which was adopted by 15+ product teams, reducing experiment bias by 30% and accelerating feature rollouts. 2. Built Data Sync pipeline with resilient aggregation, retries, and publishing to BigQuery & Kafka via Kubernetes CronJobs, processing 10M+ daily events with 99.9% reliability and reducing data latency by 25%. 3. Enhanced Experiments API, enabling dynamic filtering, date constraints, and backward compatible schema upgrades. 4. Developed Meteor’s MaxCompute metadata extractor plugin in Go during GCP→Alicloud migration, enabling complete lineage tracking for 500K+ datasets and ensuring P0 metadata capture with zero downtime. 5. Automated stale asset cleanup in Compass with custom Golang scripts, improving discovery efficiency by ∼20%, while strengthening platform reliability through Protobuf schema versioning, gRPC contract refinement, and test coverage & profiling (pprof). 6. Authored internal experimentation architecture docs (Hierarchical, Orthogonal, Switchback), driving alignment across PMs and engineering teams. 7. Used GitHub Copilot early on for auto-completion, later designing and deploying custom LLM-based agents for observability, automating log-anomaly triage and auto-generating incident summaries, cutting manual on-call investigation time.

1. Led end-to-end feature completion of In-App Campaigns for Mercury (In-house Campaign manager platform), from RFCs to scalable, fault-tolerant backend microservices. 2. Built a DB proxy with caching & query optimization, boosting system efficiency by 35%, and designed reusable APIs (Builder pattern) reducing dev effort by 15%. 3. Implemented Kafka-based async messaging and improved production stability via test refactoring, profiling (pprof), load optimization, and caching. 4. Ensured smooth campaign operations by achieving 95% critical bug resolution within 24 hours.

Optimized Golang (Echo, MongoDB, SQL) features for an Ed-tech platform, reducing server response time by 20%, improved concurrency with goroutines (15% efficiency gain), and built REST APIs with FDB (File Database) for secure learning workflows.