B
Brian Xu
About
Detail
California, United States
Timeline
work
Job
school
Education
Résumé
Jobs
verified_user
0% verified
- MSenior Software EngineerMAX.AIApr 2024 - Current (1 year 11 months)• Led the development of Mavrick V3 to a no-code orchestration engine for building production-grade multi-agent flows, combining LangChain, LangGraph, LiteLLM and Ray Serve behind a FastAPI layer. • Directed the product roadmap for Maverick V3 + Nova Agent, translating executive vision into clear engineering milestones and coordinating with design, ML, and UX teams. • Integrated retrieval-augmented generation with PostgreSQL, Redis Vector + Faiss, and OpenAI function-calling to ground agent responses on client data. • Built a multi-layer hallucination mitigation framework for Maverick by combining Guardrails and Retrieval-Grade layers, schema-constrained function calling. • Engineered dynamic pricing and billing systems using Node.
Senior Software Engineer (Backend)DatabricksNov 2022 - Apr 2024 (1 year 6 months)• Served as lead engineer on cross-org efforts connecting Delta Lake with AI/ML tooling, aligning infra APIs with platform roadmap and product requirements. • Designed cloud-agnostic infrastructure using Kubernetes and Terraform to power big data and ML workloads on Databricks, enabling scalability across AWS, Azure and GCP. • Architected multi-tenant SaaS infrastructure in AWS, ensuring secure data isolation and scalability for enterprise customers. • Collaborated closely with front-end teams to achieve seamless integration using JavaScript & TypeScript (Next.js, Vue.js and React), prioritizing user-centric services and integrated with Firebase (GCP). • Led performance optimization for Delta Sharing, reducing data transfer latency- Senior Software Engineer (Full Stack)RedditApr 2021 - Nov 2022 (1 year 8 months)• Owned critical paths in Reddit's comment delivery pipeline, improving backend reliability and latency under peak loads with a 30% reduction in rendering delays. • Built and optimized Go, Java and Python microservices for post editing, voting and reporting through optimized DB SQL queries and migration of REST to gRPC services. • Conducted the migration of the content database from a monolithic PostgreSQL schema to Cassandra clusters, improving write throughput by 2x for comments and posts. • Refactored WebSocket and polling-based update systems, reducing duplicate comment fetches by 30%, integrated Kafka and RabbitMQ to ensure reliable message delivery. • Authored backend APIs and documentation using OpenAPI (Swagger), improving
- Software Engineer (Full Stack)GoogleOct 2018 - Apr 2021 (2 years 7 months)• Led the migration of security services to a microservices-based architecture with Go, Java and Python frameworks, improving scalability and fault tolerance, supporting traffic spikes. • Shipped a Borg-hosted binary-scanner that inspected 10K + prod build per day, trimming median patch lag for critical CVEs 30% within one quarter. • Authored Sapience static-analysis rules that surfaced 1,200+ insecure crypto usages in Google3, led fixes before public disclosure. • Added a differential-privacy aggregation layer to YouTube Ads logs (Beam on Dataflow), enabling GDPR-compliant reports within 1% metric skew. • Migrated Ads log ingestion from MapReduce to Flink-on-Borg, doubling sustained throughput to 8M QPS while holding P99 latency f
- LSoftware Engineer InternLinkedIn (intern)May 2017 - Aug 2017 (4 months)• Deployed an anomaly-detection pipeline on Apache Spark Streaming and Kafka, classifying around 3B behavioral events every day and driving 35% drop in spam. • Added GraphX login-graph and geo-velocity features, boosting fraud-model AUC from and eliminating 20% of false positives. • Blended GBDT and deep auto-encoders in TensorFlow/scikit-learn with Python, surfacing 1M fake connection requests weekly at ms-level latency. • Dockerized the scorer and scrolled to 500+ prod nodes via Azkaban + Samza, achieving zero-downtime deployment in three weeks. • Authored an internal white paper – later my MIT MEng thesis – standardizing the fraud stack for LinkedIn's 2018 Trus-and-Safety playbook.
Education
verified_user
0% verified
- Master of EngineeringMassachusetts Institute of TechnologySep 2017 - Jun 2018 (10 months)
- Bachelor's Degree – Computer ScienceMassachusetts Institute of TechnologySep 2014 - May 2018 (3 years 9 months)