Looks like you have JavaScript disabled. For the full Twine experience, you will need to re-enable it.

I'm Timothy A. Kosecki, a Model Behavior Architect focused on AI trust & safety and risk-sensitive domains. I design structured evaluation frameworks, run red-team exercises, and translate behavioral findings into implementable guardrails that improve safety, reliability, and policy alignment. I bring a background in financial crime investigation and political/mis/disinformation review, applying consistent criteria to complex, ambiguous cases. I'm currently pursuing the Google Data Analytics Professional Certificate to strengthen SQL, data cleaning, exploratory analysis, and decision-ready visuals.…I'm Timothy A. Kosecki, a Model Behavior Architect focused on AI trust & safety and risk-sensitive domains. I design structured evaluation frameworks, run red-team exercises, and translate behavioral findings into implementable guardrails that improve safety, reliability, and policy alignment. I bring a background in financial crime investigation and political/mis/disinformation review, applying consistent criteria to complex, ambiguous cases. I'm currently pursuing the Google Data Analytics Professional Certificate to strengthen SQL, data cleaning, exploratory analysis, and decision-ready visuals.

Timothy A. Kosecki

Data Analyst, Audio Transcriber, Data Collector, +2





I'm Timothy A. Kosecki, a Model Behavior Architect focused on AI trust & safety and risk-sensitive domains. I design structured evaluation frameworks, run red-team exercises, and translate behavioral findings into implementable guardrails that improve safety, reliability, and policy alignment. I bring a background in financial crime investigation and political/mis/disinformation review, applying consistent criteria to complex, ambiguous cases. I'm currently pursuing the Google Data Analytics Professional Certificate to strengthen SQL, data cleaning, exploratory analysis, and decision-ready visuals.…I'm Timothy A. Kosecki, a Model Behavior Architect focused on AI trust & safety and risk-sensitive domains. I design structured evaluation frameworks, run red-team exercises, and translate behavioral findings into implementable guardrails that improve safety, reliability, and policy alignment. I bring a background in financial crime investigation and political/mis/disinformation review, applying consistent criteria to complex, ambiguous cases. I'm currently pursuing the Google Data Analytics Professional Certificate to strengthen SQL, data cleaning, exploratory analysis, and decision-ready visuals.

Available to hire

I’m Timothy A. Kosecki, a Model Behavior Architect focused on AI trust & safety and risk-sensitive domains. I design structured evaluation frameworks, run red-team exercises, and translate behavioral findings into implementable guardrails that improve safety, reliability, and policy alignment.
I bring a background in financial crime investigation and political/mis/disinformation review, applying consistent criteria to complex, ambiguous cases. I’m currently pursuing the Google Data Analytics Professional Certificate to strengthen SQL, data cleaning, exploratory analysis, and decision-ready visuals.

Skills

Experience Level

Expert

Expert

Expert

Expert

Expert

Work Experience

Model Behavior Architect at Innodata Services, LLC

July 1, 2024 - Present

Lead quality assurance and structured evaluation of agentic and generative AI systems for two Fortune 10 clients, defining repeatable evaluation criteria and focusing on answer quality, safety, reliability, and policy alignment. Design and execute red-team and adversarial test suites (jailbreaks, prompt injections, coercive escalation, misuse chaining) to uncover model vulnerabilities and inform guardrail design. Analyze edge cases and behavioral patterns, translating findings into clear policy recommendations and implementation guidance for deployment teams. Collaborate with cross-functional stakeholders to refine model behavior, enforce trust & safety standards, and document defensible decisions. Mentor peers on evaluation rigor and written rationales, improving consistency and quality across multiple high-stakes projects. Designed and iterated system prompts and multi-step agent workflows to improve answer quality, reliability, and safety across high-volume evaluation suites. Built

LLM Prompt Analyst (Contract) at Leading Conversational AI Provider

March 1, 2024 - June 1, 2024

Evaluated model prompts and responses against quality, safety, and policy criteria for next-generation conversational AI systems. Designed targeted test cases to surface behavioral inconsistencies, edge cases, and safety vulnerabilities. Produced concise, written recommendations to improve robustness, context handling, and trust & safety alignment. Analyzed model responses to assess intent, nuance, and safety alignment.

Senior Financial Crime Investigator at Kraken

October 1, 2021 - January 1, 2023

Investigated cryptocurrency fraud, money laundering, and sanctions evasion using blockchain analytics tools and SQL-driven transaction data. Analyzed transactional patterns and entity networks to identify suspicious activity and build cases for AML escalation. Drafted and filed detailed SARs/STRs and supporting risk assessments for regulators and internal stakeholders. Trained and mentored junior investigators, reinforcing analytical rigor and sound investigative judgment.

Financial Crimes Investigator (Contractor) at Cash App (via AML RightSource)

July 1, 2020 - October 1, 2021

Conducted end-to-end investigations of suspicious transactions, fraud, and potential money laundering across digital payments and cryptocurrency products, using monitoring systems and blockchain analytics tools to identify patterns. Documented findings and wrote clear case summaries and regulatory reports, ensuring accuracy and BSA/AML compliance.

Specialist (E-4), 82nd Airborne Division at U.S. Army

July 1, 2007 - March 1, 2011

Executed mission-critical operations under strict protocols in high-pressure environments. Maintained a Secret security clearance, demonstrating reliability and discretion. Coordinated with diverse teams to complete complex tasks with zero tolerance for error.