Skip to content
View simer05's full-sized avatar

Block or report simer05

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.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. 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
simer05/README.md

Hi, I'm Simerjit Kaur

MSc Artificial Intelligence candidate at Nanyang Technological University (NTU), Singapore, graduating January 2027. I build LLM systems that run on real data and check their own outputs: evaluation frameworks, guardrailed assistants and multi-agent pipelines.

Before NTU I spent 2 years at NielsenIQ as a Research Analyst and Team Lead, delivering consumer research for global brands such as Reckitt and Nestlé.

🌐 simer05.github.io · LinkedIn · Google Scholar


Selected work

SenticLens · first-author paper, SENTIRE Workshop at IEEE ICDM 2026 A production sentiment-analysis framework that audits whether LLM outputs stay consistent, and flags inconsistent ones before they reach downstream analytics. Labcast talk

Kenza Living · Technical Project Lead (part-time) Bilingual (English/Arabic) community platform, live on the App Store, Google Play and the web (React Native, Node.js, PostgreSQL on AWS). Its LLM assistant turns residents' messages into bookings and service tickets, with server-side guardrails for security and correct bookings.

Agentbox · open-source contributor A desktop app that runs many Claude Code and Codex agents in parallel, each in its own git worktree. First contributor to bring Windows support, merged upstream.

RAG grounding audit Tested 3 open-weight LLMs on 500 HotpotQA questions: reordering the same passages flipped 8–14% of answers, and models answered without evidence 26–40% of the time.

TabM++ · extended the TabM-PLR tabular deep-learning baseline; improved the matched baseline on all 5 TabReD datasets.

Hardware-aware U-Net for Stable Diffusion · 55% activation sparsity, 24% lower memory, 12.6% faster inference.

CodeLah! · LoRA-fine-tuned CodeGen-350M for code ↔ natural-language translation (+40.8% over baseline).


Hackathons

  • Sea x OpenAI Codex Hackathon, Regional Finalist · AdaptLink, AI seller intelligence for Shopee: seller dashboard and a 6-agent seller workflow.
  • American Express GenAI Hackathon, Top 6 Finalist · GenAI rewards personalisation with a Safe Action Engine (policy-as-code) and RAG-based engagement analytics.
  • TE Connectivity x AWS Hackathon, 2nd Place · TE AI Data-Center Copilot (team of 4): built the competitor and company-performance analysis and the RAG copilot on AWS (Textract, Bedrock, Knowledge Base with cited answers), and delivered the final presentation and demo.

Tools

Python · TypeScript · PyTorch · Hugging Face · LangGraph · React Native · Node.js · Next.js · PostgreSQL · Docker · AWS · Linux/HPC

Pinned Loading

  1. Sea_x_OpenAI Sea_x_OpenAI Public

    AdaptLink: AI seller intelligence for Shopee (Sea x OpenAI Codex Hackathon, Regional Finalist)

    TypeScript 1

  2. agentbox agentbox Public

    Forked from savannahfeder/agentbox

    A desktop GUI for running dozens of Claude Code and Codex agents in parallel, each in its own git worktree. Answer them like email, or use the board view. Free, open source, on the plan you already…

    JavaScript

  3. rag-grounding-audit rag-grounding-audit Public

    Behavioural audit of whether RAG answers follow their evidence: paired evidence conditions on HotpotQA across three open-weight LLMs

    Python