Four roles across
AI & finance.

Built a configurable data-quality dashboard for 2.4K+ daily accounts using Angular, replacing 12 legacy Horkos checks & XML validation files with SQL-configurable rules backed into IBM DB2 and Sybase.

Developed gap report recalculations through the dashboard via shell script triggers through Spring for AutoSys job orchestration, reducing manual remediation from ~30 minutes to a single-click workflow.

Engineered a production LangGraph deepagent with API call, vector search and NL2SQL capabilities for root-cause analysis and remediation recommendations on flagged accounts.

Worked across RL environments spanning coding, finance and travel domains for frontier model training.

Evaluated agent trajectories, graded reward signals and validated task completion across multi-turn RL rollouts.

Performed RLHF and SFT for frontier LLMs through rubric evaluations of LLM-generated code for correctness, style and best practices across A/B test variants.

Evaluated Agent-as-a-World simulations by assessing tool use, API calls & CoT reasoning in YAML files.

Verified chain-of-thought reasoning across STEM domains by checking logical consistency in intermediate steps.

Performed model stumping to identify failures and edge cases, generating adversarial prompts for frontier LLMs.