AI Systems
Agent architectures, LLM orchestration, memory, tool calling and autonomous workflows.
AI / ML ENGINEER GENERATIVE AI AGENTIC SYSTEMS
Building AI systems
that think, retrieve,
reason & act.
AI/ML Engineer focused on production-grade generative AI, autonomous agents, RAG architectures, intelligent retrieval and scalable machine-learning infrastructure.
PythonPyTorchLangGraphRAGLLMsFastAPIAWSAzureGCPDockerKubernetes
LLM
Memory
Retrieval
Agents
Tools
Planning
Evaluation
Vector DB
APIs
SYSTEM STATUSAgentic AIRetrievalMemoryEvaluationDeploymentONLINE
Engineering intelligence, not just models.
I work at the intersection of machine learning, generative AI and software engineering, turning emerging AI capabilities into scalable, observable and production-ready systems.
Retrieval Intelligence
RAG pipelines, embeddings, semantic search, hybrid retrieval and knowledge systems.
Production AI
Model serving, APIs, evaluation, observability, security and scalable infrastructure.
Building across the AI stack.
Production engineering experience spanning autonomous agents, retrieval, model infrastructure, NLP, recommendation systems and distributed ML.
AI/ML Engineer
Building autonomous software-engineering agents for repository analysis, planning, code generation, debugging, context retrieval, agent memory and structured evaluation.
PythonPyTorchLangGraphMCPFastAPIFAISSRedisAWSKubernetes
Machine Learning Engineer
Built real-time intent prediction, recommendation systems, NLP pipelines, event-driven ML services and production cloud infrastructure.
PyTorchTransformersFastAPISparkKafkaAirflowAWSSageMaker
Academic foundation in Artificial Intelligence.
Graduate-level study focused on the theory, engineering and practical application of modern artificial intelligence systems.
GRADUATE EDUCATIONM.S.
Master of Science in Artificial Intelligence
DePaul University
Advanced academic training in artificial intelligence, machine learning, intelligent systems and the engineering foundations required to build production-grade AI applications.
Artificial IntelligenceMachine LearningDeep LearningGenerative AIAI Systems
AI
Projects where research meets engineering.
A selection of systems designed around reliability, retrieval quality, execution and measurable production outcomes.
01HEALTHCARE AI / REAL-WORLD EVIDENCE
OncoDelta
An oncology evidence intelligence platform designed to make the gap between clinical-trial outcomes and real-world patient outcomes visible, traceable and actionable.
Clinical trial vs real-world outcomesORR, PFS and OS evidence comparisonDisease and subtype intelligenceSource-traceable evidenceEvidence-grounded AI assistantHEOR and RWE workflows
Next.jsSupabaseOpenAIPostgreSQLRAGClinical Data APIs
02AGENTIC AI
Autonomous Software Engineering Agent
Agentic architecture for repository understanding, hybrid context retrieval, multi-step planning, code changes, sandbox execution, tests and human approval.
Coordinator AgentHybrid RetrievalPlanningSandbox ExecutionTestingHuman Approval
PythonLangGraphMCPPostgreSQLRedisDockerKubernetesGitHub
03EXPLAINABLE AI
AI Explainability Agent
Low-latency explainability system combining model attribution, retrieval and generative AI to turn model predictions into clear, evidence-backed explanations.
<250 msp95 latency
75%cache hit
60%+token reduction
+40%acceptance
RAG explainabilitySemantic retrievalCost-aware promptingEvaluation workflow
FastAPIFAISSLightGBMSentence TransformersDVCLLMs
04AI SECURITY & PRIVACY
AshGuard AI
Privacy-first AI platform focused on detecting sensitive information, redacting protected data, scoring privacy risk and preserving verifiable audit trails.
PII DetectionData RedactionPrivacy ScoringAI GovernanceAudit TrailsFederated Learning
PythonLLMsPrivacy GuardrailsBlockchain
05MACHINE LEARNING INFRASTRUCTURE
Real-Time Anomaly Detection
Streaming anomaly-detection pipeline for production telemetry with scalable data processing, model serving and low-latency inference.
30%fewer false alarms
<50 msp95 inference
Edge telemetryKafka streamingSpark processingModel servingKubernetes deployment
KafkaSparkTorchServeDockerKubernetesMLflowAirflow
FLAGSHIP PRODUCT / ONCODELTAExplore OncoDelta
See the gap between trial promise and real-world outcomes.
OncoDelta is built around a practical oncology evidence problem: outcomes observed in controlled clinical trials do not always match outcomes seen in broader patient populations. The product makes those differences easier to compare, inspect and trace back to evidence.
COMPARETrial and real-world ORR, PFS and OS.
TRACEKeep evidence connected to source data.
RESEARCHExplore oncology evidence with grounded AI.
MONITORFollow new RWE and registry evidence.
EVIDENCE COMPARISONTrial vs Real World
+13 ptsexample outcome delta
Illustrative interface example based on the live product presentation.Inside My AI Stack
A production AI system is more than an LLM. I design the retrieval, orchestration, data, infrastructure and evaluation layers around it.
INTERFACE
WebMobileIDEAPI
AI GATEWAY
AuthenticationRate LimitingSessions
AGENT ORCHESTRATION
CoordinatorPlanningRetrievalExecutionEvaluation
INTELLIGENCE
LLMsPrompt ManagementMemoryTool CallingReasoning
RETRIEVAL
EmbeddingsVector SearchBM25Hybrid SearchReranking
DATA
PostgreSQLRedisVector DBObject Storage
INFRASTRUCTURE
DockerKubernetesAWSAzureGCP
OBSERVABILITY
OpenTelemetryPrometheusGrafanaEvaluation Pipelines
How I build AI systems.
Understand
Clarify the user, business goal and constraints.
Retrieve
Find the most relevant structured and unstructured context.
Reason
Use models and agents to plan the path to an answer or action.
Execute
Connect AI to tools, APIs, databases and workflows.
Evaluate
Measure correctness, latency, reliability, cost and safety.
Improve
Feed validated failures and feedback back into the system.
AI Engineering Stack
Tools selected for the system being built, not for a buzzword checklist.
AI / LLM
OpenAIClaudeGeminiPrompt EngineeringTool UseLLM Evaluation
Agentic AI
LangGraphLangChainCrewAIAutoGenMCPMulti-Agent Systems
Machine Learning
PythonPyTorchScikit-learnNLPRecommendation SystemsAnomaly Detection
Retrieval
RAGEmbeddingsSemantic SearchHybrid RetrievalBM25Reranking
Data & Vector
PostgreSQLRedisFAISSPineconepgvectorDynamoDB
Cloud & MLOps
AWSDockerKubernetesTerraformMLflowGitHub Actions
Backend
FastAPIREST APIsgRPCPydanticMicroservicesAsync Python
Observability
OpenTelemetryPrometheusGrafanaCloudWatchEvaluation Pipelines
Questions I keep pulling apart.
Agentic AI
Reliable planning, memory, tool use and evaluation for autonomous workflows.
AGENTS / MEMORY / EVALSRAG & Retrieval
Hybrid retrieval and reranking for higher relevance with less context noise.
RAG / BM25 / EMBEDDINGSAI + Blockchain
Where verifiable provenance and auditability can strengthen AI workflows.
AI / AUDIT / PROVENANCEAI Privacy
Detecting sensitive data and enforcing policy while preserving useful context.
PRIVACY / PII / GOVERNANCEAutonomous Software Engineering
Agents that understand repositories and learn from execution feedback.
CODE / AGENTS / SANDBOXExplainable AI
Turning model behavior into explanations people can inspect and act on.
XAI / SHAP / HUMAN FACTORSWhat good AI engineering looks like.
Reliability > Demo
AI systems must work consistently outside controlled demonstrations.
Retrieval > Bigger Prompts
Give models the right information instead of merely more information.
Evaluate Everything
Without evaluation, AI systems cannot improve systematically.
Intelligence + Infrastructure
Strong AI products require both capable models and strong engineering.
akash://ai-engineer
$ initializing ai-engineer-profile > loading agent systems... [ok] agents > loading retrieval architecture... [ok] rag > loading model infrastructure... [ok] inference > loading evaluation framework... [ok] evals > current objective: BUILD INTELLIGENT SYSTEMS THAT SHIP.
5+Years building AI & ML
20+AI / ML technologies
ProductionFocused engineering
AlwaysLearning and building
CONTACT
Have an AI problem worth solving?
Whether you are building an AI product, agent platform, intelligent retrieval system or production ML infrastructure, I would be glad to talk.