GV

GeezerVison

The World Wide Web, restored to 800 pixels of glory.

Home|About|Gaming cabinet|Archives|How it works|Starter sites

On this site

  • Sign in
  • ActionsCI/CD and automation
  • CodespacesCloud development environments
  • All Features
  • Documentation
  • Training
  • Certifications
  • Build with Copilot
  • Customize Copilot
  • Integrations
  • Trending repos
  • GitHub Sponsors
  • Developer articles
  • Enterprises

This view

  • Original page
  • New address

GeezerVison is showing github.com — rklop - Overview

@rklop rklop Follow
More
  • Overview
  • Repositories
  • Projects
  • Packages
  • Stars
@rklop
rklop
Follow
View rklop's full-sized avatar

rocky klopfenstein rklop

Follow
senior at amherst college
1 follower · 0 following
  • in/rockyklop
More
  • Overview
  • Repositories
  • Projects
  • Packages
  • Stars
rklop/README.md

Rocky Klopfenstein

AI Systems · Automated Reasoning · Backend Engineering

I'm a Computer Science and Mathematics student at Amherst College, graduating in May 2027. I build AI applications and research how to evaluate and verify the code they generate.

Previously, I worked on distributed file operations and AI-powered content workflows at Box. My research spans text-to-SQL evaluation, verified code generation, and software engineering agents.

LinkedIn · Google Scholar · Email

Selected Work

Evaluating Software Engineering Agents

Can two patches pass the same tests and still behave differently? My thesis explores this question through iterative patch generation and behavioral test synthesis, with inspectable results from a 100-instance experiment.

Text-to-SQL Generation & Verification

Combines schema-aware SQL generation with bounded equivalence checking using veriEQL. Synthesized counterexample databases help identify incorrect queries that execution on a single database can miss.

Python · SQL · SMT Solving · LLM Evaluation

Tagging the Switch

Investigates language identification and part-of-speech tagging in English-Spanish code-switched text. Compares single-task, sequential, and multi-task transformer training, including uncertainty-based loss weighting.

PyTorch · Hugging Face Transformers · Multilingual NLP

Drop AI

A full-stack application for email classification, reply generation, and PDF analysis, connecting a Flask backend with a Next.js frontend and Gemini-powered processing.

Python · Flask · Next.js · TypeScript · Gemini

Technical Toolkit

Languages: Python, Java, SQL, C/C++, TypeScript
AI & Evaluation: LangGraph, LangChain, MCP / FastMCP, LangSmith, PyTorch, SMT Solvers
Backend & Infrastructure: FastAPI, PostgreSQL, Redis, Docker, HPC

Pinned Loading

  1. SURF_2025 SURF_2025 Public

    2025 Summer Research: Formal SQL Verification on Text-2-SQL Benchmarks | Collaboration w/ Prof. Wu & Andrew T

    Python

  2. tagging-the-switch tagging-the-switch Public

    Exploring POS & LID Tagging in English-Spanish Code Switched Tweets

    Python

  3. Drop-AI Drop-AI Public

    LLM-Powered Email Assistant | Originally HackMIT 2025, 2nd Place

    TypeScript

  4. edge-detector edge-detector Public

    Image Edge Detector w/ Sobel Operators

    Java

Something went wrong, please refresh the page to try again.
If the problem persists, check the GitHub status page or contact support.

Uh oh!

There was an error while loading. Please reload this page.

You are visitor 01402503 — GeezerVison does not store the pages you look up.
Best viewed with Netscape Navigator 4.7 or Internet Explorer 6 at 800×600. No cookies required.