Skip to content
View vignesh-nagarajan-vn's full-sized avatar
🎯
Focusing
🎯
Focusing
11 followers · 11 following

Achievements

Achievement: Pair Extraordinairex2Achievement: Pull SharkAchievement: QuickdrawAchievement: YOLO

Achievements

Achievement: Pair Extraordinairex2Achievement: Pull SharkAchievement: QuickdrawAchievement: YOLO

Highlights

  • Pro

Block or report vignesh-nagarajan-vn

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
vignesh-nagarajan-vn/README.md

Vignesh Nagarajan

Texas A&M Engineering · Computer Science × Econometrics
Machine Learning | Game Theory & Decision Science | Financial Modeling

Website LinkedIn Email

Technical Roles Relevant Awards
Deloitte · Technology & Implementation Extern Mu Alpha Theta · National Scholar, 1 of 27
Qurios Ventures (acquired) · Founding Member Congressional App Challenge · National Winner
Jetson ($3M Series A) · SWE Intern National Cyber Scholar · SANS-GIAC, $3.4K
Drippy · Data Science & ML Intern CyberStart America · Top 2.8% of 35K+
STEM·E · Data Analyst Intern AMC / AreteLabs · 2× National Champion

Selected Projects

Quarry-LDR · local deep research: one GPU compresses the web into cited evidence and writes the report
An iterative plan, search, fetch, rerank, extract, synthesize, verify loop where the local GPU compresses roughly 750K tokens of scraped text into roughly 60K of evidence, then writes the report itself. Every cited sentence must survive an entailment check before render, and the default engine makes zero API calls, with Claude as an optional paid tier.
Norn · variant-interpretation copilot that drafts ACMG/AMP evidence for a human curator
Paste a genetic variant (HGVS, rsID, or locus); Norn pulls from public genomics databases, adjudicates each criterion with a two-pass Claude reviewer, then computes the classification in code. Built at the Claude Life Sciences Hackathon (1 of 500 selected from 6,000+ applicants).
SchoolPulse AI · edge-AI suite tracking a school's water, food, and energy footprint
Statistically-gated leak detection (confidence-interval slope analysis on simulated sensor data), an EfficientNet-based waste-sorting classifier (96% accuracy, deployed via TFLite), and a RAG-driven operations agent, unified in a single dashboard. Finalist at the USAII Global AI Hackathon.
Ordinance · full-stack AI/RAG backlayer for Congressional offices
Keyword-scored RAG pipeline over a curated federal corpus (EPA rules, OMB circulars, CISA directives) that returns citation-bracketed answers anchored to CFR/USC references. All inference is proxied through serverless API routes so credentials never reach the browser.

Open-Source Contributions

yfinance · three upstream fixes (v1.6.0) to the most widely used Python library for Yahoo Finance market data
  • Fixed a 4-year-old bug where 30-min price requests raised misleading unavailability errors from an internal interval substitution (PR #2900 for Issue #1029)
  • Repaired pandas 3 test-suite breakage from zoneinfo vs pytz timezone comparisons (PR #2901)
  • Stopped error messages from falsely claiming tickers are possibly delisted (PR #2903 for 5 issues incl. #2902)
qlib · two contributions to Microsoft's AI-quant investment platform
  • Fixed two long-standing FileInstrumentStorage bugs that were corrupting qlib's instrument files, adding regression tests plus a new self-contained class and 7 new functions (PR #2302 for Issue #1720)
  • Aligned the pre-commit black formatting hook with the version CI enforces (PR #2303)

Research Papers

Always open to interesting problems and collaborations.
vigneshn26@tamu.edu

Pinned Loading

  1. REFLEX REFLEX Public

    ML for endogenous OTC bond markets, where a dealer's policy reshapes future flow. Reframes learning as finding a fixed-point equilibrium stable under self-induced market adaptation.

    Python 2

  2. Quarry-LDR Quarry-LDR Public

    Local deep research: an 8 GB GPU compresses 750K web tokens to 60K evidence; Qwen3-8B writes the cross-encoder-verified report at $0.

    Python 1

  3. SchoolPulse-AI SchoolPulse-AI Public

    Suite of edge-AI tools and a web dashboard that help schools cut water, food, and energy waste by detecting leaks, sorting compost, and visualizing consumption in one conclusive dashboard.

    Jupyter Notebook 3 1

  4. Norn Norn Public

    Drafts ACMG/AMP variant classifications for a human curator: gathers genomics evidence, adjudicates each criterion with Claude, computes the label in code via ClinGen points, and shows every source.

    TypeScript 1

  5. MIRROR MIRROR Public

    Explainable radiology AI that reads chest X-ray, brain MRI, and head CT scans, then chains classification, evidence localization, and language generation into a traceable clinician-style report.

    TeX 1

  6. Ordinance Ordinance Public

    RAG-powered operating system for Congressional offices that turns legislation, ethics rules, and appropriations into "Actionable Insights" (AI), answering natural-language queries with cited sources.

    TypeScript 1