Photo of Decker Brower

Decker Brower

Senior Software Engineer · Tech Lead

I raise the floor wherever I go.

I build platforms teams can trust — by making high-impact systems reliable, observable, and safe to scale. Lately that means production AI agents: measurable, grounded, and cheap enough to run at scale. I've spent my career in startups and scale-ups, shipping product while hardening the foundations underneath it.

Now

At Handshake I lead engineering on an agentic AI recruiting assistant, which I took from proof of concept to production, alongside the ads and analytics platform behind employers' paid job promotions. The part I care about most is making AI systems measurable: every agent change gets replayed against real production traffic before it ships. Long-term, I'm focused on Staff-level impact: owning problems end-to-end, raising the bar for teams, and building systems that last.

What I do

  • Production AI agents: evals and replay harnesses, grounding, and the cost and latency work that make agents worth trusting.
  • Trustworthy metrics: clarify semantics, prevent silent inflation, ship data quality checks.
  • Debuggable systems: tracing and dashboards that make "why is p95 spiking?" a five minute answer.
  • Execution at scale: patterns + guardrails that reduce surprises without slowing teams down.

Selected impact

  • AI agents: rewrote a production LLM agent's orchestrator as deterministic code, cutting LLM cost ~59% and halving latency; grounding work took fabricated details in drafts from 20 to 0.
  • Data platform: cut a core analytics pipeline from 18.8 hours to 81 minutes, saving ~$78K/yr; cut messaging-database IOPS 75% ahead of peak traffic.
  • Games: built Volley's first 3D game, CoComelon, cloud-streamed to LG, Samsung, and Fire TV; shipped 4 new voice games, driving exponential monthly subscriber growth over the following 4 months.
  • Revenue: introduced Marquee Ads, driving a 900% surge in ad revenue; improved article ad placement to boost impressions by 70% and revenue by 37%.
  • Commerce: shipped BOGO deals, resulting in a 73% increase in AOV and a 6000% surge in order volume month-over-month for a partner.
  • Reliability: reduced speech-to-text failure rate by 99% by implementing a circuit breaker pattern across games.
  • Teams: redesigned an engineering hiring process and trained 30+ engineers to run technical and system-design interviews.

Projects

Right now it's really one thing: I'm building an AI game development platform — a tool so anyone can make the games they always wanted to play and share them with friends. GameMaker was my first taste of this as a kid; I want that spark, but easier, with AI doing the hard parts. The games are demos it built end-to-end.

AI Game Development Platform

You're the director; the AI is the studio. Describe the game you want and an agentic platform does the heavy lifting — code, art, 3D, and wiring — then ships it straight to the browser. Going from an idea to a playable, shareable game is fast and approachable.

Godot 4.7 · Blender · Python · TypeScript · Cloudflare · Terraform

Explore the platform & play the games →

 Playpen Play the games → Everything the platform has shipped, in your browser. No install.

Experience

Handshake — Senior Software Engineer, Ads & AI (Jun 2025 – Present)

Engineering lead for production AI agents and the ads and analytics platform behind employers' paid job promotions.

  • Took an agentic AI recruiting assistant from proof of concept to production in about two months, across the agent pipeline (Python, Google ADK, Cloud Run), Rails/GraphQL backend, React UX, and Terraform.
  • Rewrote its orchestrator as deterministic code: LLM calls per run 9.0 → 2.4 (~59% lower cost) and p50 latency 16.7s → 8.2s.
  • Built a PII-safe replay harness that gates every agent change on real production conversations; grounding work it validated took fabricated details from 20 to 0.
  • Tech lead for Promotions Analytics (Go, Rails, dbt/BigQuery, React), from beta with 30 employers to GA.
  • Cut a core data pipeline from 18.8 hours to 81 minutes (~$78K/yr saved) and messaging-database IOPS by 75%.
  • Rolled out workload-identity caller authentication (OIDC/JWKS) across four services, and led incident response across the ads auction and data platform.
Volley — Tech Lead, Product & Platform (Jun 2024 – Jun 2025)

Tech lead for voice-driven and cloud-streamed TV games, and for how the engineering team hired.

  • Built Volley's first 3D game, CoComelon, in Unity, streamed to LG, Samsung, and Fire TV through AWS GameLift.
  • Shipped 4 new voice-controlled AI games for Roku, driving rapid subscription growth.
  • Reduced speech-to-text failure rate by 99% via circuit breaker pattern.
  • Redesigned the engineering hiring process: new technical and system-design interviews, and trained 30+ engineers to run them.
  • Improved testing strategy to reach 90%+ test coverage on a team.
Leafly — Senior Software Engineer, Ads & Commerce (Nov 2020 – Jan 2024)

Worked across ads + commerce systems with a bias toward measurable outcomes: revenue, conversion, performance, and quality.

  • Introduced BOGO deals (+73% AOV, +6000% MoM volume for a partner).
  • Introduced Marquee Ads (+900% ad revenue).
  • Boosted ad placement performance (+70% impressions, +37% revenue).
  • Enabled native iOS checkout flows (supported 40% of annual orders).
  • Reduced accessibility errors by 94% and daily checkout errors by 95%.
  • Improved performance: jobs (-99.99% runtime), search false negatives (-99.9%), latency (-43% / -50%).
FullStack Labs — Senior Software Engineer (May 2017 – Nov 2020)
  • Led a 30-person team building a high-complexity financial accounting system for Medicaid payment processing.
  • Spearheaded a 20-person team delivering nine custom B2B web apps in a year.
Earlier roles (2012 – 2017)
  • Built and shipped a multi-sided B2B/B2C SaaS platform in freight/logistics.
  • Shipped an Android/iOS WebView market research app.
  • Built bespoke CRM + portfolio management tools.

Skills & tools

I work across product and platform engineering — Python, TypeScript, Go, Ruby, and C#; React, Rails, Node, and FastAPI; Postgres, BigQuery/dbt, Spanner, Redis, and Elasticsearch; GCP, AWS, Cloudflare, Kubernetes, and Terraform with Datadog, Grafana, and OpenTelemetry for observability. Lately I've been deep in production AI agents (Google ADK, evals, replay harnesses), voice (Deepgram, ElevenLabs), and real-time games (WebRTC, Unity, Godot).

Say hi

The fastest way to reach me is via LinkedIn DM.
The best way is an invitation for beer and tacos.