A cost-bounded LLM pipeline for content verification
A standalone service that finds landing pages that load fine but contradict the ads pointing at them — and estimates the spend at risk — built as a funnel so most pages never reach an LLM.
- Python
- FastAPI
- Playwright
- MongoDB
Builder — internal hackathon, Sep 29 – Oct 1
One symptom, three root causes
Enterprise customers were fully blocked by one error message. I traced it to three unrelated causes — missing data, a swallowed error shape, and a bad field mapping — and turned a hard failure into a self-service recovery path.
- C# / .NET
- PHP
- React
- Redis job queues
Engineer
Reconciling state after a worker dies mid-operation
A worker crashed after an external system had created entities but before we saved them — leaving users a permanently broken view and duplicates on retry. I built a recovery path that rebuilds our state from the external system, without ever claiming an entity that isn’t ours.
- C# / .NET
- MongoDB
- Meta Graph API
Engineer
An async media pipeline, from zero to production
Video creation through a third-party API takes minutes to ingest, so the existing blocking request path couldn’t support it. I designed a self-scheduling async workflow with pre-flight validation and deduplicated uploads, and shipped it across two products and their frontends.
- C# / .NET
- React
- Redux
- AWS S3
Lead engineer
Merging two diverged codebases without regressions
Two React applications had forked years earlier and drifted — ~580 shared files duplicated or diverged. I planned and executed a staged merge (355 files moved, 339 deduplicated, ~245 reconciled with LLM-assisted review), then fixed all 13 launch blockers within a week.
- React 16 / 18
- Redux
- Webpack
- i18next
Lead engineer
Filtering large datasets through a query DSL
A filter system that compiles UI state into a central query service’s DSL, with a new distinct-values endpoint, typeahead over very large sets, keyset pagination and persisted state.
- React
- Redux
- C# / .NET
- SQL
Engineer
Making LLM tool-calling reliable in production
The model is rarely the bottleneck — schemas, context routing and truncation are. I designed versioned tool schemas for an in-product assistant, fixed silent capability gaps, built multimodal generation, and rolled out search backed by vector retrieval.
- LLM tool calling
- MCP
- Prompt engineering
- Vector search
Engineer
Extending a rule engine without breaking stored rules
Turned a customer feature request into general relative-date operators for a rule engine, backward compatible with every stored rule — and extended a weather-driven automation tool to a new platform with fuzzy geo matching.
- C# / .NET
- React
- OpenWeather API
- Meta Marketing API
Engineer
Async bulk onboarding for a company-wide product merger
Core engineer on a merger with a hard business deadline. Rebuilt account onboarding from synchronous per-row writes into async queued jobs with batched inserts, and shipped the first statistically-gated automation for two new platforms.
- PHP
- C# / .NET
- Redis job queues
- OAuth 2.0
Core engineer
A hierarchical multi-tenant dashboard, POC to GA
Owned the delivery epics for a dashboard that rolls thousands of accounts up into portfolios and groups — data model, cross-service APIs, access-loss handling and a fast hierarchical UI — and extended it to new platforms without a rewrite.
- React
- Redux
- PHP
- C# / .NET
Delivery-epic owner
Solo, from PRD to a premium feature in production
The only engineer on a premium feature — I reviewed the PRD, built the POC, wrote the design doc, then implemented batch writes to a third-party API with partial-failure semantics, plan gating and analytics.
- C# / .NET
- Google Ads API
- React
- Redux
Sole engineer
Cutting 10.8 TB of storage without breaking a user feature
Two storage prefixes had grown without bound to 8 TB and 2.8 TB. I redesigned retention around lifecycle policies and moved the write path — while keeping the customer-facing download feature working.
- AWS S3
- C# / .NET
- PHP
Engineer
Making a multi-tenant pipeline survive its own scale
A daily job that processes every customer account against rate-limited third-party APIs kept failing under load. I added cross-process backoff, adaptive concurrency, per-unit fault isolation and real observability — so it degrades gracefully instead of falling over.
- C# / .NET
- Redis
- MongoDB
- AWS S3
Owner
Moving a pipeline onto a shared data-cache layer
Migrated an entire platform’s recommendation stack off direct report reads onto the shared data-cache architecture, then built new recommendations, audits and dashboard widgets on top.
- C# / .NET
- Microsoft Advertising API
- React
- PHP
Engineer
Bridging two frontend stacks with a build plugin
Wrote a Webpack plugin that renders React components inside server-rendered PHP pages, which let one navigation system replace two diverged ones across every tool.
- React
- Webpack
- PHP
- CoffeeScript
Engineer (intern → SDE)
A product built 0 → 1, shipped to every user
Designed the frontend architecture and worked on the backend generation pipeline for a new recommendation product — ~20 independent modules, virtualized lists at 60 fps, per-module fault isolation — and replaced the legacy tool for all users in 7 months, as an intern.
- React
- Redux
- C# / .NET
- AWS S3
Core engineer (intern)