Research → Action

Most academics study problems. I also build solutions.

These projects translate my research on platform capitalism, labour process, and digital alternatives into working systems that people can actually use.

⚡ Stata Support for Positron (positron-stata)

First-class Stata language support, interactive data explorer, live console execution, and variables tracking for Positron IDE.

The Problem

Stata has a legendary statistical engine and remains the primary analytical environment for empirical economics, development studies, and econometric research. However, its default user interface and do-file editor have not kept pace with modern developer workflows.

Positron (by Posit) is a next-generation data science IDE offering fast virtualized data viewing, but it previously lacked Stata integration, leaving researchers stranded between isolated windows.

What It Does

positron-stata bridges your locally installed, licensed Stata engine (Stata 17, 18, or 19 — BE, SE, and MP) directly into Positron:

  • Native .dta Data Explorer: Double-click any .dta file to open it in Positron’s high-speed virtualized Arrow data viewer. Instant smooth scrolling, sorting, and filtering over millions of rows with zero lag.
  • In-Column Sparklines & Distributions: In-header mini-histograms, unique count profiling, and missing-rate warnings right at the top of every column before estimating models.
  • Automatic Value Label Decoding: Categorical variables display their human-readable labels (e.g., Married, Single, College grad) automatically while retaining exact underlying numeric data types.
  • Interactive Do-File Ergonomics: Break do-files into interactive blocks using * %% cells or **# section headings, complete with inline ▷ Run Cell CodeLens buttons.
  • Live Console & Return Inspection: Integrated execution in the terminal, automatic tracking of e() and r() return scalars and matrices in the native Variables pane, and plot history in the Plots tab.

Rapid Adoption & Impact

  • 1,400+ organic downloads within 24 hours of release on Open VSX.
  • Brings modern IDE developer ergonomics to the empirical research and economics community without disrupting existing Stata licenses, ado-files, or institutional workflows.

Status & Links

🍽️ OpenCoop / OpenFood

A cooperative food delivery platform where everyone benefits: workers, restaurants, and customers.

The Problem

Platform capitalism extracts 25-35% commission from restaurants, suppresses worker earnings through algorithmic coercion, and hides money flows behind opaque systems.

My PhD research documented these mechanisms in detail. This project is the alternative.

What It Does

OpenCoop is a cooperative food delivery platform with:

  • 0% restaurant commission — restaurants receive 100% of menu price
  • Workers keep 80%+ of delivery fees + 100% of tips
  • Daily minimum wage guarantee (₹600, governed by cooperative vote)
  • Open job board — workers see all jobs, no algorithmic penalties
  • Transparent fees — every charge breakdown visible
  • Cooperative governance — one member, one vote

Why It Matters

This project demonstrates that cooperative platform models are technically and economically viable. It translates my PhD research on platform capitalism into a working alternative.

Technical Details

Modular monolith designed for pilot scale (~100 restaurants, ~300 workers):

  • Backend: Node.js/TypeScript, Express.js, PostgreSQL 16, Redis 7
  • Frontend: Next.js 14, Tailwind CSS
  • Key features:
    • Event-sourced architecture with append-only audit log
    • Redis job board (workers choose jobs, not algorithmic assignment)
    • SHA-256 hash chain for tamper-evident records
    • Governance-by-vote (proposal → vote → auto-execute)
    • Escrow via Stripe Connect with worker guarantee pool

Status & Links

  • Status: Active development, pre-pilot
  • GitHub: github.com/abhinavjnu/opencoop
  • License: Open source (commons-owned)
  • Tech stack: Node.js, TypeScript, PostgreSQL, Redis, Next.js, Docker
  • Documentation: Full architecture docs in repo

📊 IndiaData-colab

Automated analysis pipeline for Indian government survey data, helping researchers with limited compute bypass the need for massive infrastructure.

The Problem

Large-scale Indian survey datasets (PLFS, NSS, HCES, ASI) are:

  • Massive (hundreds of thousands of records)
  • Complex (require specialized knowledge of survey weights, strata, clusters)
  • Difficult to analyze without expensive local compute or R/Python expertise
  • Prone to errors if multipliers and sampling weights aren’t handled correctly

This creates a barrier for researchers, especially those without institutional resources or technical training.

The Solution

Zero-installation analysis:

  • Google Colab notebook — click badge, upload CSV, run → get results
  • No local R setup needed — runs entirely in browser
  • Validated against real data — 100% exact match to official MoSPI calculations (< 0.01pp)

Official MoSPI methodologies:

  • Principal + Subsidiary Status (PS+SS) — official “Usual Status” approach
  • Current Weekly Status (CWS) — short-term labour status
  • Principal Status (PS) — strict majority-time status
  • Automatic weight computation — handles NO_QTR, combined subsamples, Calendar Year datasets

Key features:

  • Labour indicators: LFPR, WPR, UR (overall + disaggregated)
  • Intelligent survey design with correct weights, strata, clusters
  • Export to Word (.docx) and LaTeX
  • Memory-efficient Parquet export for large datasets
  • API integration with microdata.gov.in for data discovery/download

What This Demonstrates

  • Public good tool for Indian labour market research community
  • Lowers barriers to rigorous survey data analysis
  • Transparent, open-source alternative to proprietary statistical software
  • Enables replication and verification of published studies

Validation

Tested on real MoSPI PLFS Calendar Year 2024 data (415,000+ records):

Indicator Manual Pipeline Diff
LFPR (CWS) 55.2% 55.2% 0.01pp
WPR (CWS) 52.4% 52.4% 0.01pp
UR (CWS) 5.0% 5.0% 0.00pp
LFPR (PS+SS) 59.6% 59.6% 0.01pp
WPR (PS+SS) 57.7% 57.7% 0.01pp
UR (PS+SS) 3.2% 3.2% 0.00pp

✅ ALL INDICATORS MATCH MANUAL GROUND TRUTH (< 1pp)

Status & Links

  • Status: Active development, pre-pilot
  • GitHub: github.com/abhinavjnu/IndiaData-colab
  • Colab Notebook: Open in Colab
  • Tech stack: R, Python, Google Colab, data.table, survey, srvyr
  • License: Open source
  • Datasets supported: PLFS, NSS, HCES, ASI (any microdata.gov.in dataset)

Contact

Interested in collaborating on these projects? Reach out:

Email: maurya.abhinava@gmail.com