Software engineer & ML fairness researcher

ArchitRathod

I build systems that are measurably correct and demonstrably fair — from OpenStreetMap road closures at city scale to fairness auditing for deep models.

Selected work ↓Résumé ↗GitHub ↗

01 — Selected work

Field surveys

01

FairLint DL↗

Deep-learning fairness debugger shipped as a VS Code extension

AI/ML · Open Source2026
02

OSM Temporary Road Closures↗

Database & API for the OpenStreetMap Foundation · GSoC 2025

Full Stack · OSS2025
03

Freight Toolkit↗

CMAP regional freight dashboard · UIC-UTC × FTA FERSC

Full Stack · Urban Analytics2025
04

relfair↗

Relationship-aware counterfactual fairness testing · PyPI

AI/ML · Open Source2026
05

FairGuard↗

Agentic investigative journalism — mapping federal revolving doors

AI Agents · Responsible AI2026
06

TransitLake↗

Dagster-orchestrated medallion lakehouse over Chicago transit

Data Engineering · Geospatial2026

All 39 projects ↗

02 — Research

Published coordinates

ASE 2026

FairLint-DL

FairLint-DL is a Visual Studio Code extension that brings fairness debugging directly into the developer workflow before training completes. The tool trains a configurable deep neural network as a proxy model, then applies information-theoretic Quantitative Individual Discrimination (QID) metrics grounded in Shannon and min-entropy to detect bias. Its two-phase gradient-guided search discovers discriminatory instances, while a causal debugging pipeline localizes bias to specific layers and neurons using sensitivity analysis. Dual explainability engines based on SHAP and LIME provide feature-level attribution. On the Adult Census Income dataset, the system reports that 96.0% of analyzed instances exceed the 0.1-bit QID threshold, with a mean QID of 0.619 bits and a disparate impact ratio of 0.581, and it produces these results within 12 seconds on cached models.

2026 · 2 authorsASE 2026

ICDSA 2024

Ascend.AI

The proposed approach embarks on an intricate and comprehensive exploration of advanced and innovative technologies to enhance interview skills. Leveraging the power of OpenCV and Xception, the paper delves into the nuances of facial expression analysis, unraveling the intricacies of emotion recognition. The system analyzes tone and pitch with the aid of tools like LIBROSA to extract vocal features in order to understand the intensity of emotions, and has developed a 1D-CNN model for classification using RAVDESS, TESS, SAVEE, and CREMA-D datasets. The system includes a chatbot using vector database Qdrant and an open-source LLM Mixtral 8x7b, offering personalized interview guidance derived from scraping 30 diverse websites for interview-related questions. This technical exploration extends from conventional interview preparation to introducing an innovative framework that intertwines machine learning models with real-time analysis.

2024 · 4 authorsSpringer Nature Singapore

NeurIPS 2023

Multiagent Social Simulators

Multiagent social network simulations are an avenue that can bridge the communication gap between the public and private platforms in order to develop solutions to a complex array of issues relating to online safety. While there are significant challenges relating to the scale of multiagent simulations, efficient learning from observational and interventional data to accurately model micro and macro-level emergent effects, there are equally promising opportunities — not least with the advent of large language models that provide an expressive approximation of user behavior. In this position paper, we review prior art relating to social network simulation, highlighting challenges and opportunities for future work exploring multiagent security using agent-based models of social networks.

2023 · 8 authorsarXiv ↗

All 9 papers ↗

03 — The surveyor

Engineer by trade,
cartographer by instinct

I finished my MS in Computer Science at the University of Illinois Chicago in May 2026, where my research sat at the intersection of algorithmic fairness, explainability, and geospatial systems — auditing the models we trust and mapping the cities we live in.

Recent waypoints: a Google Summer of Code project for OpenStreetMap, a freight analytics platform for 285+ Chicago municipalities, a NeurIPS workshop paper on multiagent social simulators, and relfair — a PyPI library for relationship-aware counterfactual fairness testing.

Full route history ↗

9

Publications & reports

39

Projects charted

285+

Municipalities served

8.5M+

OD pairs processed

103,481

Mortgage applications audited

5+

Years shipping

04 — Claude Code

But what are skills without Agent Skills?

A collection of custom Claude Code skills I use daily — with copyable SKILL.md files so you can drop them straight into your own setup.

All 9 skills ↗

Design

/frontend-design

Create distinctive, production-grade frontend interfaces with high design quality. Generates creative, polished code that avoids generic AI aesthetics.

SKILL.md ⧉

Productivity

/skill-creator

Create new skills, modify and improve existing skills, and measure skill performance with eval-driven iteration loops.

SKILL.md ⧉

DevOps

/claude-md-improver

Audit and improve CLAUDE.md files in repositories. Scans for all CLAUDE.md files, evaluates quality, outputs a graded report, then makes targeted updates.

SKILL.md ⧉

Research

/session-report

Generate an explorable HTML report of Claude Code session usage — tokens, cache hits, subagents, skills, and expensive prompts — from local transcripts.

SKILL.md ⧉