Skip to content
SS
Press the power button on the remote to start the Shikhar Sisodia workspace.
FinderFileEdit
shikhar/experiments/research.md

experiments / research.md

Research

Five works across model evaluation, automation, astrophysics, structural engineering and economics — with the record type and contribution role stated for each.

01research projectOngoing

Evaluating LLM Antidepressant-Prescribing Predictions Against Real EHR Data

Author · Shikhar Sisodia

Undergraduate research at Santa Clara University testing whether three open models reproduce real antidepressant prescribing distributions. Model output is mapped into six drug classes and compared against OCHIN electronic health record data across demographic subgroups.

model output vs. real prescribing distributions

450 groups · 25 runs each · 3 prompt tiers

SSRIsSNRIsNDRIsTCAsNaSSAsOthers← output mapped into six classes
  • LLaMA 3.1 8B
  • Mistral 7B
  • Mistral-Nemo 12B
TOP-1hit rateMAElower is betterJSDlower is better
LLaMA 3.1 8B
top-1 90.4% · mae 0.159 · jsd 0.427
Mistral 7B
top-1 90.4% · mae 0.169 · jsd 0.467
Mistral-Nemo 12B
top-1 90.2% · mae 0.159 · jsd 0.429

Plotted with better upward in every panel. The top two tie on Top-1 and swap order on MAE and JSD — the lines cross — so no single model leads outright. Bias is analysed separately across race, gender, federal poverty level.

Spearman correlation · Top-1 hit rate · MAE · Jensen-Shannon divergence · benchmarked against OCHIN electronic health records

This evaluates prediction of prescribing patterns in aggregate. It is not medical advice and not a prescribing system.

02data paperDec 11, 2025

N8N LinkedIn Creators Agent: A Workflow for Discovering AI Builders at Scale

Researcher · Shikhar Sisodia, Abyan Kashif, Ryan Saleh, Aryan Goel

A modular automation workflow for discovering LinkedIn creators who publicly share AI-agent and workflow builds. Framed as a repeatable discovery process rather than a ranking or recommendation system.

ZenodoSSRNdoi 10.5281/zenodo.17982978

discovery workflow

a repeatable process, not a ranking system

  1. 1AI-assisted query generationgenerate
  2. 2Public Google searchretrieve
  3. 3Structured result extractionextract
  4. 4Email-pattern inferenceinfer
  5. 5Persistent loggingpersist

The workflow ends at persistent logging. No discovered profiles are shown here.

03reportSep 8, 2025

The Great Silence: An Experimental Exploration of the Fermi Paradox and Its Implications for Astrobiology

Co-author · Rutu Tahasildar, Shikhar Sisodia

Examines the paradox's historical context and its quantitative framing through the Drake Equation, groups the proposed resolutions into categories, and connects them to the observational programmes that test them.

Zenododoi 10.5281/zenodo.17484161

drake equation · declining certainty

term constraints as the paper states them — no value is computed

N =R∗ · fp · ne · fl · fi · fc · L
R∗fpneflfifcL
Each point is the paper's own characterisation of how well constrained that term is.
  • constrained
  • varies
  • wide range
  • unknown
  • R∗rate of star formation≈1.5–3 stars/year
    constrained · Chomiuk & Fomalont, 2012
  • fpfraction of stars with planetsclose to 1
    constrained · Cassan et al., 2012
  • nehabitable planets per systemestimates vary wildly
    varies · Kopparapu, 2013
  • flfraction where life arisesinfinitesimal to almost certain
    wide range · Tirard et al., 2010
  • fifraction developing intelligenceunknown
    unknown · Shubin, 2008
  • fcfraction that broadcastmay choose silence
    unknown · Tarter, 2007
  • Lhow long they stay detectablemost critical and uncertain
    unknown · Ashworth, 2015

Certainty collapses from left to right. That is the paper's point — the equation frames the uncertainty rather than resolving it, so no N is computed here.

  • Rarity of intelligent life
  • Undetectability or disinterest of advanced civilisations
  • Our own temporal position in cosmic history
  • SETI
  • Exoplanet research
  • Astrobiology

04working paperMar 27, 2025

Origami-Based Kinetic and Radiative Shielding for Space Habitats

Co-author · Rutu Tahasildar, Shikhar Sisodia

Investigates a Yoshimura-crease structure as a deployable dual-purpose shield: Finite Element Analysis of its response to hypervelocity micrometeoroid and debris impact, paired with a theoretical model of its deployed volume as a container for radiation-shielding material.

Zenododoi 10.5281/zenodo.17479927

yoshimura lattice · 16 circumferential × 8 axial

drawn to the paper's stated geometry

stoweddeployed 2 m
diameter
1 m
deployed
2 m
wall
3 mm
material
Aluminum 6061

Kinetic

Static stress simulation in Autodesk Fusion 360 with non-linear material properties, base fixed. Reports von Mises stress, displacement, deformation — a qualitative read of how the folds distribute impact energy, not a rated effectiveness.

Radiative

Double-walled: stowed volume is negligible, deployed volume is filled with water (1 g/cm³). The metric is areal density (g/cm²)— added mass over the cylinder's surface area.

05data paperDec 19, 2025

Cryptocurrency and Financial Inclusion

Editorial contribution · Ashmit Sethi

A mixed-methods study of Bitcoin adoption and financial inclusion, combining literature review, policy analysis and media reporting with correlation, regression and predictive analysis of price and transaction-volume data, and Bitcoin ATM availability relative to population.

Zenododoi 10.5281/zenodo.17986137

mixed-methods design · el salvador

methodology and named analysis scripts

~70%

of Salvadorans lack access to a conventional bank account

  • no conventional bank account
  • banked

methods

  • Literature review
  • Policy analysis
  • Media reporting
  • Correlation & regression
  • Predictive analysis
  • ATM availability vs. population

analysis scripts

  • atm.pyATMs against city population
  • graphsall.pyprice and volume visualisations
  • method1.pyregression against SME growth and credit access
  • method2.pycomparative and predictive analysis

Findings concern adoption, small businesses, volatility, infrastructure and financial stability. No results series is published, so none is charted here.

research identity

A persistent identifier for the researcher. It is not a publication, and the works above are listed from their own deposit records rather than from an ORCID work list.

  • Algorithmic bias
  • Astrobiology
  • Clinical decision support
  • Cryptocurrency
  • Deployable structures
  • Drake Equation
  • Financial inclusion
  • Finite Element Analysis
  • Information retrieval
  • LLM evaluation
  • Mixed methods
  • SETI
  • Space habitats
  • Workflow automation