Skip to content
Production-minded Naperville, Illinois

01 / Senior AI Software Engineer

Building production AI systems that reason, retrieve, speak, automate, and scale.

I design and ship AI agents, voice systems, retrieval platforms, recommendation engines, and full-stack products from architecture through production.

Explore my work View resume LinkedIn
AI Systems/Full-Stack/ML Infrastructure
agent-topology / livemulti-provider
YearsSoftware engineering
AI & automation workflows
AI voice calls / month
Manual work reduced
Service availability
RAG accuracy improvement

02 / ENGINEERING PHILOSOPHY

Engineering AI Beyond the Demo

Production AI is a systems problem. I build the product, data, orchestration, infrastructure, evaluation, and operational controls that make model capability dependable for real users.

architecture → evaluation → operations
  1. 01

    Production First

    AI features must survive real users, latency constraints, failures, permissions, changing data, and operational complexity.

    reliability / latency / safety

  2. 02

    Systems Thinking

    The model is one component in an architecture spanning retrieval, tools, events, APIs, observability, security, and human escalation.

    models inside dependable systems

  3. 03

    Measurable Quality

    Quality is evaluated with real datasets, test conversations, ranking metrics, latency, cost, reliability, and business outcomes.

    evaluate / observe / improve

  4. 04

    Full-Stack Ownership

    Move comfortably between product interfaces, backend services, model orchestration, databases, infrastructure, and deployment.

    interface to infrastructure

03 / Selected systems

Production systems, not isolated demos

Selected systems built across Daniel’s professional work—designed around real users, operational constraints, measurable quality, and long-running ownership.

Featured production systems

System / 01

production

AI Revenue Operations Platform

Agents · Automation · Multi-tenant SaaS

A multi-tenant AI platform coordinating voice agents, chat workflows, CRM automation, scheduling, retrieval, and customer operations across 18 workspaces and more than 75 configurable workflows.

client workspaces
18
configurable workflows
75+
less repetitive work
61%
response time
<3 min
down from 26 minutes
ReactNext.jsFastAPILangGraphOpenAIAnthropicPostgreSQL+3
revenue-ops / control-plane18 workspaces

System / 02

production

Realtime Voice Agent Platform

Voice AI · Realtime Systems · Provider Routing

Production voice automation supporting lead qualification, scheduling, customer support, outbound campaigns, transfers, and after-hours workflows.

calls / month
~4,500
provider failover
Automatic
interruption handling
Realtime
CRM-aware conversations
Contextual
VapiDeepgramOpenAIAnthropicElevenLabsTwilioAmazon Transcribe+1
voice-runtime / streamingprovider failover

System / 03

production

Permission-Aware RAG Platform

Retrieval · Search · Evaluation

A retrieval platform grounding AI responses in customer websites, policy documents, pricing data, campaign history, and internal sales documentation.

grounded-answer accuracy improvement
24%
pgvectorOpenSearchOpenAI embeddingsFastAPIPostgreSQL
rag-pipeline / permission-awaregrounded

System / 04

production

Marketplace Search & Recommendation Platform

Ranking · Recommendations · Forecasting

Machine-learning-powered product discovery and recommendation infrastructure for a wholesale marketplace.

recommendation CTR
+15%
average order value
+7%
zero-result searches
−22%
forecasting error
−14%
PyTorchXGBoostAmazon PersonalizeElasticsearchOpenSearchPineconeOpenAI embeddings+1
discovery / rankingsignal blend

04 / Reference architecture

How I build AI systems

A production AI product is a chain of explicit layers. Each one has a different failure mode, latency budget, and operational responsibility.

Select a layer to inspect its runtime role. The architecture is intentionally provider-aware without making the product provider-dependent.

How I build AI systems
production-ai / reference-architectureinstrumented

Layer 03

AI Orchestration

State, tools, routing, and policy turn model calls into controlled workflows.

Runtime components

LangGraphAgentsModel RouterGuardrails
traceroute selected
path03 / 08

Explicit boundaries

Synchronous product paths stay separate from retriable asynchronous work.

Controlled action

Permissions, guardrails, approval gates, and escalation surround model behavior.

Inspectable quality

Application traces and AI evaluations share the same operational picture.

05 / Experience

Built layer by layer

A career grounded in full-stack delivery, expanded through production ML and search, and focused now on complete AI systems.

Professional experience
Chapter 03

Proven ROI

Senior AI Software Engineer

— Austin, Texas

Owned the architecture and production discipline behind multi-tenant agents, voice automation, retrieval, CRM workflows, and the infrastructure used to operate them.

75+
workflows
~4,500
calls / month
+24%
grounded accuracy
−61%
repetitive work
  • Led architecture and delivery of a multi-tenant AI revenue-operations platform built with React, Next.js, FastAPI, LangGraph, PostgreSQL, pgvector, Redis, and AWS, supporting 18 client workspaces and 75+ configurable AI and automation workflows.
  • Built inbound and outbound voice agents for lead qualification, scheduling, customer support, campaign follow-up, and after-hours coverage, processing approximately 4,500 calls per month through Vapi, Deepgram, OpenAI, Anthropic, ElevenLabs, and Twilio.
  • Introduced a provider-independent speech and model layer across Deepgram, Amazon Transcribe, ElevenLabs, Amazon Polly, OpenAI, Anthropic, and Amazon Bedrock, routing by latency, quality, availability, and cost with automatic failover.
  • Connected voice and chat agents with HubSpot, Salesforce, and GoHighLevel for customer context, lead scoring, CRM updates, meeting scheduling, confirmations, and qualified-call transfers.
  • Improved grounded-answer accuracy by 24% with permission-aware RAG over websites, pricing, policies, campaign history, and sales documentation using pgvector, OpenSearch, metadata filtering, reranking, citations, and confidence-based escalation.
  • Reworked lead-management automation around EventBridge, SQS, Lambda, n8n, idempotent APIs, retry policies, audit logs, and approval gates, reducing repetitive work by 61% and average response time from 26 minutes to under three minutes.
  • Standardized deployment, observability, and AI evaluation with Docker, ECS Fargate, Terraform, GitHub Actions, CloudWatch, OpenTelemetry, Langfuse, and Sentry; built 450+ test conversations, added safeguards for high-risk behavior, and mentored four engineers.

Core stack

ReactNext.jsFastAPILangGraphPostgreSQLpgvectorRedisOpenSearchOpenAIAnthropicAmazon BedrockVapiDeepgramElevenLabsTwilioAWSDockerECS FargateTerraformOpenTelemetryLangfuse
Chapter 02

Faire

AI/ML Software Engineer

— San Francisco, California

Built marketplace product systems, then applied ranking, search, recommendation, catalog intelligence, and forecasting models at production scale.

+15%
recommendation CTR
−22%
zero-result searches
−14%
forecast error
99.9%
availability
  • Delivered supplier onboarding, catalog management, wholesale ordering, retailer accounts, payments, inventory visibility, and order tracking with React, TypeScript, FastAPI, PostgreSQL, Redis, and AWS for approximately 9,000 retailers, 1,200 suppliers, and 620,000 listings.
  • Designed personalized recommendation services with collaborative filtering, PyTorch ranking models, product embeddings, behavioral data, and Amazon Personalize experiments, increasing recommendation click-through rate by 15% and average order value by 7%.
  • Rebuilt product discovery with Elasticsearch, Amazon OpenSearch Service, synonym expansion, behavioral ranking signals, learning-to-rank, Pinecone, and OpenAI embeddings, reducing zero-result searches by 22%.
  • Automated product classification, attribute extraction, supplier-data normalization, duplicate detection, and description generation across approximately 140,000 monthly catalog updates, cutting manual review effort by 46%.
  • Trained XGBoost and neural-network forecasting models using order history, seasonality, promotions, pricing, supplier lead times, and category trends, reducing forecast error by 14% and stockouts by 17%.
  • Productionized ML and generative-AI services with SageMaker, FastAPI, ECS, Lambda, Airflow, Step Functions, Docker, MLflow, and CloudWatch, maintaining 99.9% availability and improving inference latency by 36%; also delivered a GPT-3.5 Turbo support pilot that raised schema-valid outputs from 71% to 88%.

Core stack

ReactTypeScriptPythonFastAPIPostgreSQLRedisPyTorchXGBoostAmazon PersonalizeElasticsearchOpenSearchPineconeOpenAI embeddingsSageMakerAWSAirflowMLflow
Chapter 01

KitelyTech

Full-Stack Software Engineer

— Chicago, Illinois

Established the full-stack, API, cloud, and reliability fundamentals behind responsive SaaS products and multi-tenant business systems.

11
products shipped
2,400+
business users
−41%
API response time
−29%
production incidents
  • Shipped 11 responsive web applications and SaaS products with React, TypeScript, Node.js, Python, Django, PostgreSQL, and AWS for professional services, retail, logistics, and business operations clients.
  • Created 35+ reusable REST API endpoints for authentication, accounts, payments, scheduling, document processing, notifications, and reporting; integrated Stripe, Salesforce, Twilio, Google Maps, and logistics platforms to accelerate client integrations and reduce development effort.
  • Developed a multi-tenant operations platform with role-based access control, configurable workflows, reporting dashboards, audit histories, and customer self-service capabilities for more than 2,400 business users.
  • Modernized four legacy applications into modular cloud-hosted services using Docker, AWS EC2, RDS, S3, Nginx, automated database migrations, and GitHub Actions, increasing deployment frequency from biweekly to three or four releases per week.
  • Improved performance and reliability through Jest and Pytest automation, SQL optimization, caching, structured logging, monitoring, and root-cause analysis, reducing average API response time by 41% and production incidents by 29%.

Core stack

ReactTypeScriptNode.jsPythonDjangoPostgreSQLAWSDockerEC2RDSS3GitHub ActionsJestPytest

06 / Capabilities

Technology organized around outcomes

The stack follows the system: interfaces, services, intelligence, data, infrastructure, and the controls required to keep it reliable.

Technical capabilities

Stateful agents, tool use, routing, guardrails, and dependable structured behavior.

OpenAI APIAnthropic ClaudeAmazon BedrockOpenAI Agents SDKLangGraphLangChainFunction CallingMCPStructured OutputsAgent OrchestrationPrompt EngineeringContext EngineeringModel EvaluationFine-tuning

07 / AI ecosystem

One system, many disciplines

Production AI sits at the intersection of models, retrieval, interaction design, software architecture, infrastructure, and operational feedback.

AI technology ecosystem

Production AI Systems / Agents

Stateful reasoning, tool use, structured outputs, and guardrails.

LangGraphOpenAI Agents SDKFunction CallingMCP

08 / Measured outcomes

Engineering impact

Quality is only useful when it changes the operating reality: better answers, faster paths, fewer failures, and stronger product outcomes.

Observed across AI, ML, search, APIs, and operations

Selected engineering impact
metric / 01

reduction in repetitive work

Event-driven lead-management automation

metric / 02

improvement in grounded-answer accuracy

Permission-aware retrieval and reranking

metric / 03

increase in recommendation CTR

Personalized marketplace ranking

metric / 04

reduction in zero-result searches

Hybrid product discovery

metric / 05

lower ML inference latency

Production ML and generative-AI services

metric / 06

faster API response time

Query optimization, caching, and service tuning

metric / 07

fewer production incidents

Testing, monitoring, and root-cause analysis

metric / 08

service availability

Production ML services

09 / Education

Academic foundation

University of Chicago

Bachelor’s Degree

Dates attended
to
Location
Chicago, Illinois

10 / Contact

Production systems · End-to-end ownership

Let’s build something that has to work in production.

I’m interested in engineering teams working on AI-native products, intelligent automation, developer infrastructure, voice AI, retrieval systems, and complex software products.

Naperville, Illinois
Email MeLinkedInDownload Resume

danielkparker94@gmail.com