Nick Lackman
Staff Software Engineer · Distributed systems & AIBremerton · Washington

Distributed systems.
AI in
production.

I’m Nick, a Staff Software Engineer building the platform layer other engineers build their agents on: agent runtimes, real-time voice infrastructure, and the evaluation that tells a team whether an agent is actually any good. My work sits at the intersection of distributed systems and AI, where the design meets production.

See the work
26,000 requests / day

Across SMS, chat, email, and voice

~6,000 calls / day

Production leasing AI voice product I helped take to GA

−689 ms

Substantive-turn latency in a controlled VAD study · p < 0.001

13+ years

From founding engineer to shared platforms and agent infrastructure

01 / Selected work

The platform.
The systems it grew from.

Shared agent infrastructure, the production systems that proved it out, and the evals that make quality visible.

02Real-time agent systemsVoice / Streaming / Latency

A voice agent is
a systems problem.

I built Agentix, the orchestration framework for RealPage’s production leasing AI voice agent, and led the product from prototype to GA. It currently handles about 6,000 calls a day, with 400,000+ units onboarded.

Profiling and optimizing the full audio pipeline cut median response time from about five seconds to about three seconds. In a separate 100-turn controlled study, local VAD reduced substantive-turn latency by 689 ms (22%; p < 0.001, Cohen’s d = 1.04).

Integration experience

  • OrchestrationLangGraph · OpenAI Agents SDK
  • Real-time interactionOpenAI Realtime API
  • Telephony & transportWebSockets · WebRTC · SIP · gRPC
  • Audio handlingOpus · μ-law · PCM16
03EvaluationEvals / Quality signals

Evals for output
with no ground truth.

Voice output has no ground truth, so I originated reference-free audio evals to measure its quality. DNSMOS, turn-taking measures, and audio LLM-as-judge now run in production.

Public reference implementation: github.com/lackmannicholas/voice-evals

Three evaluation layers

  • Layer 1Acoustic & perceptual quality
  • Layer 2Conversational dynamics
  • Layer 3Audio LLM-as-judge

Reference-free voice quality signals.

04Developer infrastructureTerraform / Provisioning

Infrastructure that
unblocks teams.

At RealPage, I managed engineering teams and restructured Terraform provisioning, reducing new environment creation from six months to one day across 190 engineers.

That work sits underneath the application: giving teams the environments they need to build, test, and ship their systems.

Provisioning impact

  • BeforeSix months per environment
  • AfterOne day per environment
  • Reach190 engineers
  • FoundationTerraform provisioning

02 / Writing

Notes from
production.

Distributed systems, AI, and lessons from running them in production. All writing ↗

03 / Experience

Hands-on depth.
Cross-team perspective.

Individual contribution, informed by having led engineering teams, a platform group, and a startup’s engineering function.

13+

years in software engineering

BBA in Management Information Systems, Washington State University; MS in Computer Information Systems, Boston University.

RealPage · Oct 2024–Present

Staff Software Engineer – Agentic AI

Agent Platform and real-time voice

  • Co-architect and co-own the internal Agent Platform. My work focuses on agent identity, deployments, and the voice gateway.
  • Designed the code-first SDK path that lets engineers build with platform primitives from their own repositories.
  • Led the leasing AI voice product from prototype to GA; it handles about 6,000 calls daily with 400,000+ units onboarded.
  • Built Agentix, then assisted the migration to the OpenAI Agents SDK.
  • Originated reference-free audio evals, now running in production
  • Designed the path for agents to run on Google Enterprise Agent Platform or Team-hosted as an alternatives to the internal runtime service.
  • Led Leasing agent GA readiness with the security and AI safety teams, securing Architecture Board clearance.
  • Implemented and tuned guardrails to fit the voice latency budget.
  • Led the build and deployment of an MCP server exposing Knock APIs.
  • Advised on the evals platform and on Leasing AI and Resident AI performance at GA.
  • Led caching and performance work, ElastiCache migration, deployment processes, and LangSmith operations.
RealPage · Oct 2022–Oct 2024

Senior Software Engineering Manager

Telephony, Auth, Data Services, IaC

Managed engineering teams and restructured Terraform provisioning, reducing new environment creation from six months to one day across 190 engineers.

Knock · Jan–Oct 2022

Staff Software Engineer, then Software Engineering Manager

Staff: Jan–Jul 2022 · Manager: Jul–Oct 2022 · Telephony, Auth, Data Services

Led a Python 2.7-to-3 migration spanning 600,000 lines of code, 800 endpoints, and 50 asynchronous Celery tasks.

OceanSmart · Aug 2019–Jan 2022

Founding Engineer & Engineering Manager

Maritime software platform

One of five founding tech leads. Built engineering teams and helped take the company to profitability in year two.

Oceaneering Global Data Solutions · Aug 2014–Aug 2019; CGI · Mar 2013–Aug 2014

Software Developer II; Software Developer

Early engineering roles

Built production systems before moving into engineering leadership and agent infrastructure.

04 / Contact

Let’s talk.

Agent platforms, real-time systems, and the evaluation problems in between.

What I want to do next

Own a core agent primitive.
Make it useful across teams.
Keep improving it in production.