I design and build data systems, AI runtimes, and consumer products. Most of what I make lives at the boundary between infrastructure and experience — where the shape of a system determines what becomes possible for the people using it.
I'm a software engineer and founder working across the data and startup ecosystem. My work tends to sit at the seams — between raw infrastructure and polished product, between AI capability and human workflow, between real-time systems and the quiet reliability people expect from things that just work.
I build in Rust, TypeScript, and Swift. I think in systems, ship in sprints, and care most about the moment where a hard technical idea becomes something a person can actually use. I'm drawn to problems that require both engineering depth and creative instinct — the kind where architecture decisions shape user behavior, and design choices have infrastructure consequences.
A single-binary data ingestion engine written in Rust. Skippr discovers schemas, ingests from sources like S3, Kinesis, and databases, writes through a durable WAL, and compacts into Snappy Parquet on S3 — registering Glue tables and partitions for Athena. Exactly-once semantics, schema evolution, dead-letter routing. No cluster to operate.
A WebSocket-native agent runtime that connects language models to real data infrastructure. Suites define tool registries, policies, and prompts — the data engineering suite plans, authors dbt models, validates against live warehouses, and publishes. Strict JSON actions, human-in-the-loop interrupts, multi-tenant thread persistence, and pluggable backends across Snowflake, BigQuery, Postgres, and Athena.
A platform that combines a declarative rules engine with structured questionnaires for legal due diligence. Workflows bind versioned decision models to rich content definitions — branching on answers, entity types, and lender rules. Immutable published versions, evidence collection, completion tracking, and a clear handoff path to human review.
A dating app built end-to-end — discovery, matching, real-time chat, feed, push notifications, in-app purchases, deep linking. Native iOS and Android with Cognito auth, WebSocket-driven state, APNs and FCM push, and a serverless backend on Lambda and DynamoDB.
An investing app that rewards accumulating knowledge. Tiered monthly reward pools, XP-weighted distribution, gamified financial education — designed to make learning about markets feel like progress, not homework. Regulatory-safe framing as promotional incentive, not investment return.
Root-cause resolution for cloud infrastructure. CloudWatch alarms flow through SNS into a real-time dashboard where engineers acknowledge, silence, and resolve incidents. DynamoDB Streams push state over WebSocket so every team member sees the same picture. Multi-tenant, Cognito-secured, deployed with CDK.
Ingestion pipelines, schema inference, WAL-based durability, Parquet compaction, exactly-once delivery. Building the plumbing that makes analytics possible.
LLM orchestration, tool-use loops, strict action parsing, human-in-the-loop policies, vector search. Agents that operate on real systems, not toy demos.
WebSocket protocols, DynamoDB Streams, SNS fan-out, push notifications, live state synchronisation. The infrastructure behind things that feel instant.
AWS-native architecture — CDK, Lambda, multi-account pipelines, cross-region deployments. Infrastructure defined in code, deployed without ceremony.
React Native apps with real-time features, in-app purchases, deep linking, native push. End-to-end: from the store listing to the Lambda behind it.
Performance-critical runtimes, data engines, and agent loops. Single-binary deployments, compile-time safety, and the kind of reliability you forget is running.