Applied AI Systems
- AI-assisted self-service workflows
- LLM classification and structured decision support
- Automated routing, escalation, and tool-driven integrations
- Human-in-the-loop workflows with structured outputs
Maksym Prus
Senior Software EngineerAI, Automation & Product Platforms
Building scalable operational platforms, AI-assisted workflows, voice systems, and customer-facing products across proptech and enterprise environments.
$ whoami --engineering
About
Senior Software Engineer with 15+ years of experience building production web platforms, operational automation, AI-enabled workflows, and business-critical systems.
Experienced across backend architecture, modern web applications, voice and telephony integrations, and end-to-end product delivery. Current work includes applied AI, self-service automation, speech processing, provider and tenant workflows, mobile applications, and internal operational platforms.
I focus on pragmatic architecture, maintainable systems, clear operational ownership, and using AI where it creates measurable product or business value.
$ systems --applied-ai
AI & Automation
Applied AI is useful when it is connected to a dependable product workflow: validated inputs, explicit tools, observable decisions, and a safe path to human review.
Voice & Speech
- Speech-to-text and call-processing pipelines
- Text-to-speech and telephony integrations
- Call transcription and operational processing
- Conversational experiences connected to real workflows
Product Automation
- Maintenance, support, and provider operations
- Tenant journeys and self-service guidance
- Compliance, inspection, and onboarding flows
- Internal tools that replace repetitive manual coordination
Production Engineering
- Failure handling, fallbacks, and safe rollout paths
- Cost, latency, and reliability trade-offs
- Observability, analytics, and operational ownership
- Integration with existing products and business systems
Supporting context
- OpenAI
- Claude
- Whisper
- Deepgram
- Google TTS
- Voximplant
$ career --ownership
Experience
A progression from software delivery and backend ownership to full-stack product work, platform responsibility, and applied AI automation.
Senior Software Engineer
Product and platform engineering across AI-assisted operations, customer journeys, and internal workflows in proptech.
- Build and evolve AI-assisted workflows across tenant, provider, maintenance, and operational journeys.
- Integrate LLMs, speech recognition, text-to-speech, and telephony into production products.
- Deliver backend and frontend systems using PHP, Laravel, Vue, Nuxt, TypeScript, PostgreSQL, Redis, Docker, and AWS.
- Design automation for maintenance, inspections, compliance, move-outs, onboarding, and provider operations.
- Contribute across architecture, implementation, production support, analytics, and product delivery.
- Work with product and operations teams to convert manual workflows into scalable software.
Applied AILaravelVue & NuxtPostgreSQLAWS
Senior Software Engineer
Sigma Software Group
Delivered full-stack platform capabilities for an e-commerce intelligence product serving complex research and data workflows.
- Evolved backend services and customer-facing product flows across an established production platform.
- Balanced feature delivery with maintainability, integration constraints, and operational support.
- Collaborated across engineering and product disciplines to turn data-heavy requirements into usable software.
Backend ownershipNode.jsYiiProduct delivery
Senior Full-Stack Web Developer
ApexTech Inc.
Built and maintained customer-facing applications and enterprise systems across varied product environments.
- Delivered end-to-end features spanning backend services, web interfaces, and system integrations.
- Delivered software across PHP, JavaScript, React, Vue, and Node.js codebases with different lifecycle constraints.
- Took responsibility for implementation quality, technical trade-offs, and reliable software delivery.
PHPReactVue.jsNode.jsFull-stack delivery
Earlier Engineering Experience
eDoctrina · Itera Research · EmblazeOne · Cyber Relief
Progressed from PHP development into senior backend and full-stack responsibilities across education, enterprise, and web-product environments.
- Built production web applications, APIs, data-backed workflows, and browser interfaces.
- Developed broad delivery experience across legacy systems, framework migrations, and evolving product requirements.
PHPYii & Yii2JavaScriptMySQLLegacy modernization
$ architecture --model
Engineering Playground
Explore how architecture decisions affect latency, reliability, capacity, and cost under different traffic patterns. The AI workflow adds automation and human-review trade-offs. Outputs are bounded illustrative estimates, not measured production benchmarks or capacity promises.
estimate.jsonSteady API
- Throughput capacity
- 2,122 req/s
- Estimated p95 latency
- 86 ms
- Reliability model
- 99.56%
- Capacity headroom
- 41%
- Relative cost index
- 4.5×
CDNLoad balancerStateless APIPostgreSQLRedis cache
Trade-off: Simple operations and strong consistency; database scaling becomes the main constraint.
$ git show --stat engineering-decisions
Case Studies
How This Portfolio Was Engineered
A transparent look at the engineering decisions behind this site. It is not a customer case study and includes no invented production metrics, dates, or repository claims.
Make experience inspectable
A static profile and logo inventory described skills, but gave visitors little evidence of how architectural trade-offs, interaction quality, and progressive enhancement are handled.
Preserve the running system
- Keep the Pages Router, React 18, local fonts, existing routes, and easter eggs.
- Add no service, remote data source, or npm dependency.
- Support keyboard navigation, reduced motion, and narrow screens.
Layer capabilities progressively
Stable anchors form the navigation contract. Native controls provide the baseline; deterministic local models and optional transitions enhance it without owning the page.
Evidence over decoration
Visitors can explore bounded models, inspect trade-offs, review delivery context, and filter a deliberately prioritized stack.
AI-Assisted Operational Workflow
An anonymized implementation pattern based on production engineering experience. No confidential company data or proprietary workflows are shown.
Turn unstructured requests into safe actions
Operational teams need to classify incoming requests, gather context, and trigger the correct next step without hiding uncertainty behind a model response.
Keep control at system boundaries
- Integrate with existing production systems and variable-quality inputs.
- Respect cost and latency limits while preserving reliability and auditability.
- Protect confidential business data and keep human review available.
Design the workflow, not just the prompt
- Use structured intake, deterministic validation, and AI-assisted classification.
- Enrich context only as required, then invoke schema-constrained tool actions.
- Apply confidence thresholds, human fallback, and operational analytics.
Practical automation with a safe exit
- Faster routing and more consistent handling.
- Reduced manual processing and improved self-service.
- Better operational visibility without hiding ambiguous cases.
- modelOutput → sideEffect
+ validate → retrieve → model → schema → confidence
+ route(confident ? automation : humanReview)
+ observe(latency, fallback, escalation, cost)