AI-First Systems
I build reliable AI systems using prompt engineering, structured workflows, and LLM evaluation frameworks.
What I Do
Prompt Architecture
• model-agnostic structured prompting
• workflows for complex LLM behaviors
• agentic buildingLLM Systems
• prompt learning
• tool-using agents
• evaluation pipelinesAI / ML Consulting
• AI-native product design
• LLM reliability
• safety & guardrails
Selected Work
AI Systems at Scale
• Created prompt architecture for AI discovery and routing systems, including system prompts & dynamic user-context injection, improving tool routing reliability, reducing hallucination-risk scenarios, & strengthening “what should I do next?” user guidanceLLM Knowledge Systems
• Built production-ready hallucination mitigation agent that filtered for 97-98% of false outputs (one week before Google DeepMind)Model Evaluations
• Developed a proprietary, in-house 100-unit model evaluation suite to stress-test a 2-part system: generalized intelligence and enterprise/market outputsAI Translation/Transcreation
• Devised a GenAI translation system spanning 42+ languages enabling dynamic/formal equivalence, localization, & regional nuance to deliver near-native accuracy for global enterprise contextsLLM Behavior
• Synthesized cross-surface prompt contracts defining global vs. local prompting responsibilities, improving system coherence, preventing duplication or prompt drift, and enabling consistent AI behavior across product experiencesHumanization
• Conceptualized an 18-part multi-shot prompt based on rigorous linguistic research (toward the asymptotic limits of human syntax, morphology, and semantics) to create heightened anthropomorphization for natural-sounding, "human"-like product surfaces
Skills
Core Skills
• Prompt Engineering for Production LLMs
• AI Agent Systems
• Zero-Shot & Few-Shot Prompting
• Prompt Architecture & Scaffolding
• Evaluation Frameworks
• AI Translation & Multilingual Systems
• Hallucination Mitigation
• LLM GuardrailsTech Stack
• LLMS: OpenAI, Anthropic
• Prompting: System prompts, multi-step chains, structured prompting
• Orchestration Concepts: tool use, function calling, agent workflows, memory
• Evaluation: A/B testing, eval suites, failure mode analysis
• Programming: Python (eval pipelines), XML, JSON
• Interfaces: API integration, prompt testing environmentsLanguages
• Persian (native / fluent)
• Portuguese (fluent)
• Spanish (fluent)
• Italian (advanced)
• Modern Standard Arabic (advanced)
• Levantine Spoken Arabic (advanced)
• French (intermediate)
• Dakota (beginner)
• Latin (reading)
• Hebrew (limited)
Relevant Academic Background
• Trained in interpretive reasoning and ambiguity analysis (PhD), applied to structuring and evaluating LLM behavior• Developed frameworks for meaning, translation, and contextual nuance, informing multilingual and cross-domain prompt design