Event-driven, serverless data pipeline to process million-line codebases across monorepos and multi-repo architectures.

Analysis augmented with AI code intelligence system to trace cross-service invocations, implicit API contracts, and dependency chains in developer workflows

Scaled at peak to $10,000 ARR in direct sales

Scope as large as 16 engineers over 3 teams as a mentor and technical leader.

Successfully executed major performance initiatives including migration of code and tooling from CJS to ESM to reduce AWS Lambda cold start impacts on latency by over 50% (tree-shaking during bundling), and development of code-splitting and CDN caching process for our microfrontends, which dramatically improved FCP, LCP, and FID.

Planned and led a multi-month side-by-side migration for our data layer from Prisma 1 -> Prisma 2, requiring massive and incremental code changes, culminating in a no-downtime removal of the old database infrastructure against a hard deadline of Prisma 1 deprecation.

Identified several issues in the open source dependencies that we used and contributed fixes to several of them.

Designed and implemented industrial materials specification and attribute management system to improve the agility of our listings process and reduce complexity associated with ops and updates to the materials that Reibus could sell.

... and much more

Conceived of and led the development of an internal AI agent to interpret email-based sourcing requests (using industry vernacular) and automate the creation of RFQs and identification of matching marketplace material, leading to a 20% reduction in time spent on data entry tasks for sales.

Migrated several services to be more event-driven to facilitate AI integration and reactivity.

Stood up AWS infrastructure for Kafka, Schema Registry, and other components for event bus to handle streaming updates for domain objects

Developed and shipped v1 Kotlin and Python Kafka domain field broker event bus client libraries for other teams to produce/consume domain object events with virtually no additional configuration

Implemented ECS services for event filter and fanout, including poison pill protection

Incorporated additional metrics for monitoring and observability, which brought the team closer to useful on-call alerting

Kotlin, Python, gRPC, Kafka, Spring, Dropwizard

AWS, ECS

Terraform, Cloudformation, Codepipeline, Consul

Responsible for developing data collection or generation opportunities and efficient labeling techniques (typically using Python or Scala). Includes identifying vulnerable source applications in GitHub, integrating ANTLR parsers and constructing customised language grammars, and engineering extraction embedding techniques over proprietary ASTs as part of a highly parallel Google App Engine microservice workflow.

Engineered, deployed, and supported the initial release of the Praetorian Diana continuous security platform (Java Spring backend, React front end) including source control and tracking integrations and the end-to-end vulnerability detection pipeline.

Green-fielded Kubernetes-orchestrated workflow for ingestion and processing of source repositories into graphs to train hierarchical graph neural networks. Architected and implemented entirely new source translation functionality to support new JVM constructs such as lambdas and method references introduced in Java 8 for use in IBM WALA-based static analysis. Contributed several bugfixes to the points-to analysis of WALA CAst.

Continuously improving and expanding the scope of the vulnerability detection model in an Agile setting. This includes identifying current model weaknesses, proposing strategies to mitigate or resolve those weaknesses, prioritising these on the basis of cost-benefit, time, parallelisability, or otherwise, and executing -- iteratively.

Own a portion of the team Agile process: standup and retrospectives.