# New York Life Software Engineering Internship

Akash Dubey worked as a Software Engineering Intern at New York Life from May 26 through August 6, 2026, focusing primarily on applied AI engineering.

## Demonstrated work

- Designed an asynchronous claims-assistance prototype across web, phone, and SMS.
- Demonstrated public questions, service procedures, authentication transitions, claim intake, form submission, and document handling across web and phone.
- Built trust-focused skills, citations, scoped data access, evaluation, failure recovery, human escalation, and graceful call completion.
- Built authorized-source knowledge workflows and read-only developer discovery tools.
- Helped pilot a design-to-code workflow with implementation, preview, and visual verification.

This page distinguishes built and tested work from pilots and incomplete exploration. It does not claim production deployment, ROI, cost savings, or customer outcomes.

Software Engineering Intern · May 26-August 6, 2026

Applied AI, built for real workflows.

At New York Life, I focused on turning agent concepts into demonstrable, testable systems—from multi-channel claims assistance to knowledge and developer workflows.

Primary focus

Applied AI engineering with careful trust, access, and evaluation boundaries.

Core work

What I built.

Multi-channel claims assistance

Designed a claims assistant across web, phone, and SMS, covering questions, claim intake, forms, and document upload. Built mock integrations and an admin test console since production data was off-limits.

Trust and evaluation

Added citations, scoped data access, persona-based and LLM-as-judge testing, safe failure handling, and human escalation.

Knowledge workflows

Turned authorized enterprise documents into linked topic pages that refresh daily and cite their sources.

Developer discovery tools

Built a local AI-searchable knowledge base and read-only tools for a large rules codebase, then trained three developers on it.

Capstone pilot

A skills-based design-to-code loop.

Agents turn a design and a ticket into tested code, then check the preview against the design.

Step 1

A design agent combined a design, work ticket, and existing-code context into an implementation plan.

Step 2

A coding agent implemented, tested, and opened the work in a local preview.

Step 3

A verification loop compared the preview with the design and iterated on visual differences.

Controlled test observations

~30 min

Simple compositions, with 95-100% measured visual accuracy.

60-70 min

More difficult compositions in the pilot tests.

Pilot test results, not production metrics.

Additional contributions

Business-analysis copilot

Helped build a Copilot agent that drafts requirements and user stories.

Document processing

Helped design a pipeline for hard-to-parse forms.

AI enablement

Wrote an AI kickstart guide and trained colleagues across teams.

Everything here was built and demoed or pilot-tested. It doesn't claim production deployment or business impact, and it leaves out private details.