Senior Software Engineer
Richard Lietz
I bring deep technical range and genuine curiosity to engineering problems that create lasting impact — distributed backends, CI/CD that actually gets out of the way, and AI systems that keep humans in the loop.
About
Deep technical range, and a bias toward problems where the impact outlasts the sprint.
Self-driven, comfortable with autonomy, and equally willing to take on the unglamorous problems. I find a way through even when the path is not obvious.
The work I care about tends to sit where infrastructure meets product: pipelines that stop costing engineers twenty minutes at a time, service scaffolding that turns a sprint of setup into an afternoon, and AI systems built so a human stays accountable for the output.
Focus
- Systems architecture for distributed backends
- RAG pipelines and agentic system design
- Kubernetes, AWS, and CI/CD at organization scale
- Security research and threat modeling
- Human-in-the-loop AI on AWS Bedrock
Experience
Senior Software Engineer · Centene Corporation
Remote · 2022 — present- Re-architected GitLab CI/CD pipelines org-wide — compile times dropped 90%, from 20 minutes to 2.
- Built reusable Go libraries for Kubernetes microservice bootstrapping, cutting setup from sprints to days.
- Planned and executed the migration of 30+ Kubernetes microservices from on-prem to AWS.
- Architected human-in-the-loop AI on AWS Bedrock to fully automate healthcare form population.
- Built an embedded business-logic debugger with dynamic links for BUAT/UAT self-service triage.
Lead Software Engineer · Kovar Designs
Easton, PA · 2016 — 2022- Built an HTTP proxy leasing system scaling to 100,000+ concurrent IP leases, with full support tooling.
- Delivered full-stack web apps in React, Angular, Vue, Go, and Java across multiple industries.
- Sole architect, lead developer, and technical advisor across all client engagements.
East Stroudsburg University · Computer Science
East Stroudsburg, PA · 2016Selected work
- 2025 — present
Human-in-the-Loop Form Intelligence
AWS Bedrock foundation models handling structured healthcare data workflows end to end, with humans as reviewers only so compliance and auditability survive. Retains context between runs and self-corrects without manual intervention. Deliberately built on Amazon Nova 2 Lite — base-model visual capability turned out to be enough, and the high-end models unnecessary.
- AWS Bedrock
- Amazon Nova
- Go
- Healthcare
- 2024
Llama 3 ZIP Code Intelligence
Hackathon project: fine-tuned Llama 3 on USGS, USPS, and Census data to model Oklahoma ZIP code relationships. Scoring algorithms and benchmarks written from scratch — no off-the-shelf metrics.
- Llama 3
- Fine-tuning
- Python
- Benchmarks
- 2016 — 2022
Proxy Leasing Platform
HTTP/HTTPS proxy leasing system scaling past 100,000 concurrent IP leases, plus the client tooling and support surface around it.
- Go
- Networking
- Infrastructure
- 2026
Elastic Session Gateway
A protocol-aware edge proxy that terminates and authenticates player sessions, then hot-swaps the backend underneath a live connection without dropping it. Backends are provisioned on demand as Fargate tasks, discovered again after a gateway restart, and retired once idle — so capacity follows real demand instead of a fixed fleet.
- Go
- Protocol design
- AWS ECS
- Connection migration
- 2026
Server-Side World Engine
A server-authoritative world runtime: custom chunk generation, region selection and bulk block operations, and a course lifecycle that saves, resets, and restores player-built regions. Block writes are batched to keep tick time flat under load, and world state persists to object storage.
- Java
- World generation
- Mixins
- Object storage
- 2024
Market Replay Laboratory
Records live gRPC market streams to disk, then replays them deterministically against a configurable strategy so a change can be judged on identical history rather than on a lucky session. Includes stream benchmarking and a desktop UI for stepping through a run.
- Go
- gRPC
- Deterministic replay
- Backtesting
- 2024
Autonomous Execution Engine
An always-on execution service driven from chat: strategies start and stop by command, positions and balance history persist in Redis, and every fill, error, and drawdown is reported back to the operator with rendered charts instead of raw logs.
- Go
- Redis
- ChatOps
- Event-driven
- 2024
Protocol Instruction Decoder
Subscribes to a high-volume transaction firehose and decodes raw program instructions into typed domain events in real time, fanning them out to position tracking, execution, and alerting through a broadcaster mesh built for backpressure.
- Go
- Binary decoding
- Streaming
- Concurrency
Stack
- Languages
- GoJavaTypeScriptJavaScriptPythonSQL
- Frontend
- ReactNext.jsAngularVueTailwind CSS
- Backend & data
- REST APIsgRPCKafkaRedisPostgreSQLMongoDBVector search
- Cloud & DevOps
- AWSKubernetesDockerGitLab CI/CDMicroservices
- AI / ML
- RAG pipelinesAgentic system designAWS BedrockLLM fine-tuningHuman-in-the-loop systemsEvaluation & benchmark design
- Architecture & security
- Systems architectureDistributed system designProtocol designSecurity researchThreat modeling
Contact
richard@kovar.coBest reached by email. If it's about work, a paragraph on the problem beats a paragraph on the process.