Product oriented Software Engineer with an ability to work through the entire stack.
- Forward Deployed Engineer: embed with the people doing the work, find the bottleneck, ship the workflow, stay through adoption.
- Solutions Engineer: earn trust with a technical buyer, show the product in their problem, isolate the failure, ship the fix.
- Product Engineer: start from who the product is for, isolate the class of failure, ship the smallest fix that protects the end user.
- Founding Engineer: own discovery through deployment on a small team, multiple hats.
- Backend: production APIs, data pipelines, and infrastructure.
- Backend and AI workflows (Python: FastAPI, Django, Flask; JavaScript: Node.js; LangChain; multi-agent systems)
- Data pipelines and migrations (AWS Step Functions, ETL, RDS, Snowflake, PostgreSQL)
- Infrastructure (Terraform, GitLab CI/CD, cloud cost control)
- Built multi-agent AI APIs on unstructured data, enabling automated reporting, note-taking, and queryable insights for internal consultants
- Designed the Snowflake-to-RDS migration using AWS Lambda and Step Functions, with cursor-based pagination and dependency-aware ordering
- Optimized PostgreSQL (connection pooling, indexing) after moving relational reporting off the warehouse, lowering AWS spend by six figures annually
- Built a Terraform-managed disaster recovery system using cross-region AWS Backup replication, with RTO/RPO defined and no always-on standby cost
- Standardized CI/CD across teams using GitLab runners and Terraform provisioning
- Built and published containerized pipeline modules: weekly financial advice by SMS, AI unit-test generation, and AI school workflows
- Presented these use cases, drove adoption on my team at Blackstone to orchestrate pipelines faster, and was selected as a Dagger Commander
- Isolated a class of setup failures behind corporate proxies and shipped the docs fix upstream. Followed with a mental-model issue and docs so gated environments stop configuring the laptop and failing before the pipeline runs
Offline-first AI education system using local LLMs, vector embeddings, and distributed content pipelines.
- Built device auth, content distribution, AI tutors, and the embeddings pipeline that turns Drive PDFs into retrievable vector content
- Linux, Shell, Python, Go, React, PySide6, Electron, SQL
- Capped how many clients can sit on a shared server at once, and cleaned up quiet connections so abandoned sessions do not fill the box
- Built an interactive visualization of why a local database lock breaks shared access, and what the shared server changes
- Tested the product by running warehouse simulations. Found a gap: if you kill the client mid-query, the heavy query keeps running on the server. This work led to a contract with GizmoData
- Two experiments could look like one in analytics because the app only sent the human-readable name
- Teams thought turning on a sync setting would keep flags up to date. It did not. Shipped an opt-in so analytics can tell experiments apart, and docs so implementors stop assuming the cache refreshes itself
LangChain multi-agent system orchestrated on AWS (Lambda, Step Functions, SQS). Automated marketing workflows for 3 small businesses in New York.