Ronak
Patel.

Michigan (Remote)

• ·Designed and deployed an internal, GenAI-powered maps and route-optimization tool (Python, React, LangGraph, Databricks) that transforms natural language requests into multi-factor, ranked dealer visit itineraries, enabling executives to strengthen relationships with critical clients.
• ·Built an LLM dealer prioritization model against a deterministic baseline via a custom eval harness, driving an evidence-based decision for replacement.
• ·Integrated Databricks Unity Catalog SQL and Google Maps APIs with caching, batched calls, rate limiting and per-call cost telemetry.
• ·Innovated an algorithm to reduce scaling of Google Maps API costs per client from quadratic to near-linear, increasing algorithm capacity and reducing costs by 60–80%.
• ·Designed an automated workflow for classifying and extracting fields of data in bulk from customer documents for development of internal models, and optimized queries, reducing runtimes.

CodePath AI301

• ·Diagnosed a stale-cache UI bug in GitLab's Vue/Apollo AI Catalog, tracing it to a CACHE_AND_NETWORK fetch policy; authored a 6-file fix with Jest coverage across two rounds of maintainer review.
• ·Authored internal analytics metric definitions for two untracked AI Catalog events in GitLab's Rails monolith, validated via Service Ping counters on a local GDK instance.

Remote (Contract)

• ·Utilized Reinforcement Learning from Human Feedback (RLHF) to help machine learning models learn more efficiently.
• ·Ensured models adhered to legal and ethical obligations (harmlessness, truthfulness, safety, etc.).
• ·Performed prompt engineering with various parameters and produced assessments of results.
• ·Performed quality assurance on other analysts' work (prompts, feedback, conversations, instruction-following, etc.).