Farid Zakaria
Santa Cruz, CA
916.849.6418 ● farid.m.zakaria@gmail.com
PROFESSIONAL EXPERIENCE
CONFLUENT 2024 - Present
Principal Engineer I (2024-2024)
- Leading the Developer Productivity organization. Our mission is to make developer lives happier and to increase velocity.
- Migrating Kafka & Flink from Gradle to Bazel;
GOOGLE 2019 - 2024
Engineering Manager II (2024-2024)
- Leading the StableHLO team within OpenXLA. Our mission is to develop a ubiquitous backward-compatible MLIR opset, inspired by HLO, under OpenXLA OSS governance.
- Enable PyTorch, JAX, and TensorFlow to target StableHLO IR
- Enable XLA and community backends to consume StableHLO programs
- Promote StableHLO within the open-source community and drive third-party contributions to promote adoption.
Engineering Manager II (2021-2024)
- Leading the Model Development team at Looker, consisting of 27 engineers across three teams. Our mission subsumed Model Semantics, including ownership of Looker's language (LookML) definition, the overall software development lifecycle for customers and engineers, the query pipeline, and compilation to over 50 dialects of SQL.
- Actively grew team, mentored engineers and engineering managers, oversaw engineering excellence programs, and shaped the organization's culture.
- Oversaw the delivery of multiple projects: SQL interface for Looker that exposes new SQL primitives for measures and exposes rich Looker API semantics to traditional JDBC clients, removal of NFS from query path, and optimizations for GCP BI Engine (BigQuery).
- Improved BigQuery P95 latency by 1.5x (~18s -> 12s) for queries with time zones
- Dropped median rendering time for benchmark dashboard from 20s -> 10s.
Engineering Manager I (2020-2021)
- Engineering leader for two teams:
- Model Semantics: In addition to stewarding LookML and its target SQL dialects, led the development of the backend, which was rewritten to use Apache Calcite to formalize the query in relational algebra, allowing simpler transformations & optimizations such as aggregate awareness.
- Model Transform: Responsible for Looker’s PDT subsystem, a powerful ETL tool used by over 80% of Looker customers that also enables query caching.
- Delivered support for BigQuery Materialized Views and customer-facing API for PDTs to enable integration with Apache Airflow.
- Rolled out a complete rewrite of the LookML compiler to address project longevity and performance (~36x speedup)
Senior/Staff Engineer (2019-2020)
- Technical lead on the Model Transform team
- Designed & implemented parallelization of ETL backend
- Enabled support for CloudTracing (OpenTelemetry)
- Tracing improved search performance 5x
- Caught & fixed numerous N+1 bugs
- Formalized a reproducible developer environment using Nix
- Designed and delivered incremental PDT creation.
- The third most requested feature for the broader Model group
- Customers cited reduced queries from 150TB to 10GB (x15000 improvement) & 18 minute builds down to less than a minute
ORACLE 2016 - 2019
Principal Member of Technical Staff
- Founding member of the Oracle Cloud Infrastructure's Object Storage team.
- Technical lead and implemented much of the web service frontend
- Designed and implemented the initial implementation of object metadata sharding
- Designed and implemented encryption at rest for customer data & on-demand streaming
- Designed and implemented offline object validation (TiB scale) to find bit–rot
- Designed & implemented a software-defined networking solution based on eBPF that performs cloud data center network tunneling at 25 Gbps to route encapsulated (MPLSoUDP) traffic via a load balancing system based on Google's Maglev whitepaper.
- Led emergency remediations for fundamental cloud infrastructure outages meeting tight customer-determined SLOs, in commercial and FedRAMP data centers.
- Implemented custom telemetry integration for Grafana.
APPLE 2015 - 2016
AMAZON & AMAZON WEB SERVICES 2012 - 2015
EDUCATION
UNIVERSITY OF CALIFORNIA, SANTA CRUZ 2021 - 2025
Doctor of Philosophy – PhD, Computer Science & Engineering
Publications
Mapping Out the HPC Dependency Chaos (SuperComputing 2022)
sqlelf: a SQL-centric Approach to ELF Analysis
Symbol Resolution MatRs: Make it Fast and Observable with Stable Linking
UNIVERSITY OF WATERLOO 2006 - 2011
Bachelor of Applied Sciences, Software Engineering
PROJECTS & ACTIVITIES
- SHRINKWRAP – A tool that embosses the needed dependencies on the top-level ELF executable
- SQELF - Explore ELF objects through the power of SQL
- Talks: The Impact of AI on Cloud Native Engineering (DockerCon 2023), Rethinking basic primitives for store-based systems (NixCon 2022), Mapping Out the HPC Dependency Chaos (SuperComputing 2022), Challenges with Java in a hermetic world (PackagingCon 2021) & Nix in the Java ecosystem (NixCon 2020)
- Communities: Active member of NixOS and an avid open-source contributor
LINKS