Skip to content
View jouzia's full-sized avatar

Block or report jouzia

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
jouzia/README.md

Shaik Jouzia Afreen H animated profile banner

Shaik Jouzia Afreen H

BCA Student · Applied AI Research · Software Engineering

I investigate how AI systems behave, build software around practical problems, and document what the evidence actually supports.

GitHub · LinkedIn · Portfolio


About

I'm Shaik Jouzia Afreen H, a BCA student at St. Joseph's College of Arts & Science (Autonomous), India.

My GitHub work is organized into two tracks:

  • Research: empirical studies of retrieval-augmented generation (RAG), LLM routing and transfer, and the water impact of AI infrastructure.
  • Engineering projects: AI applications, agent-based experiences, developer tools, and backend systems.

I value reproducible experiments, clear limitations, and honest reporting. Ongoing work is labelled as ongoing; preliminary findings are not presented as established results.


Research

These repositories document independent research projects. They are not presented as peer-reviewed publications.

1. RAG Appended-Context Study

Effects of Appended Distractor Context on RAG Answer Quality and Generation Latency

Studies how adding distractor chunks to retrieved context affects answer-quality measures and generation-call latency across BM25, dense, and hybrid retrieval pipelines.

  • Research focus: retrieval-augmented generation, distractor context, answer quality, latency.
  • Approach: controlled evaluation across retrieval approaches and distractor conditions.
  • Status: exploratory empirical study; conclusions are limited to the tested benchmark, implementation, top-k setting, and noise conditions.
  • Important scope note: this is an appended-context evaluation, not a universal claim about retrieval-ranking robustness.

2. AI Waste Heat to Water

From AI Waste Heat to Water: A Net-Water-Impact Framework for Sustainable AI Infrastructure

Develops a framework for assessing when AI infrastructure waste heat might support thermal desalination while accounting for cooling, energy, and water burdens.

  • Research focus: AI infrastructure sustainability, thermal desalination, net freshwater impact, sensitivity and uncertainty analysis.
  • Approach: transparent system boundaries and water/energy accounting that distinguish gross freshwater production from net impact.
  • Status: ongoing modeling and research framework.
  • Important scope note: modeled and illustrative outputs are not proof of real-world water savings.

3. LLM Router Transfer Study

Evaluating Cross-Benchmark Transfer of a Lightweight LLM Router

Investigates whether a lightweight prompt-feature router generalizes across benchmark families and how its results compare with fixed-model baselines.

  • Research focus: LLM routing, cross-benchmark transfer, baseline selection, evaluation reliability.
  • Approach: empirical comparison of a lightweight router and fixed-model baselines, including held-out benchmark evaluation.
  • Status: experimental results available; broader generalization and algorithmic novelty are not established.
  • Important scope note: current evidence does not establish consistent superiority over the global baseline.

Engineering Projects

These projects focus on implementing AI applications, developer tools, and software systems. Their maturity and runtime verification vary by repository; see each project's README for its current status.

AI Applications and Agent Systems

A multi-agent educational AI platform exploring agent routing, stateful study environments, conversational memory, and reward-driven evaluation, with a Next.js interface and Python components.

A conversational learning assistant experiment using Streamlit, LangChain, Gemini/Groq models, and ChromaDB, with mentor and rival-style AI personas.

An interactive study dashboard concept with a logic-learning workspace, AI companion, provisional CGPA projections, and an animated interface.

An MCP-oriented web perception and browser automation project exploring Next.js, TypeScript, Python, Notte, and Patchright.

Developer Tools and Backend Systems

A pre-implementation concept for a developer portal combining project tracking, grounded debugging assistance, and deployment telemetry. The public repository currently contains the concept README only; no working application or tests are claimed.

An event-driven notification infrastructure project using FastAPI, PostgreSQL, Redis, and Celery, with asynchronous delivery, retries, rate limiting, and observability planned in the implementation. Check the repository's testing notes for runtime-verification status.

Web Development

A portfolio website built with React, Vite, Tailwind CSS, Framer Motion, and Supabase, with authentication and content-management features.

A Next.js experiment focused on component-based web development and modern UI implementation.


Research Interests

  • Retrieval-augmented generation and information retrieval
  • Robustness to distractor context and retrieval defects
  • LLM routing, model selection, and cross-benchmark transfer
  • Evaluation design, baselines, uncertainty, and reproducibility
  • Efficient AI systems and inference cost
  • Sustainability and net-water impact of AI infrastructure

Technical Toolkit

Languages: Python · JavaScript · TypeScript · Java · HTML · CSS

Frontend: React · Next.js · Vite · Tailwind CSS · Framer Motion

AI and data: LLM APIs · Gemini · Groq/Llama · LangChain · AI agents · ChromaDB · Streamlit

Backend and infrastructure: FastAPI · PostgreSQL · Supabase · Redis · Celery · Vercel

Tools: Git · GitHub · GitHub Actions · VS Code · Linux

Technologies listed reflect project work and ongoing learning; proficiency and runtime verification vary by project.


Achievements

  • 1st Place — PPT Presentation: won first place in a presentation competition; no certificate was issued.
  • 3rd Place — Start-up Business Plan Competition: intra-departmental competition at St. Joseph's College, 2 August 2024.
  • QuizOff 2026 — India's Biggest AI Quiz: recognition/participation associated with CampusCrew and Unstop, 19 July 2026.

Certifications

  • IBM SkillsBuild — Classifying Data Using IBM Granite
  • HubSpot Academy — Social Media Certified; valid 23 May 2026 – 21 June 2028
  • CampusCrew × Unstop — QuizOff 2026: India's Biggest AI Quiz; recognition/participation certificate, 19 July 2026

Education

Bachelor of Computer Applications (BCA)
St. Joseph's College of Arts & Science (Autonomous), India


GitHub Activity

Shaik Jouzia Afreen H GitHub statistics

Shaik Jouzia Afreen H top programming languages

Contributions

GitHub contribution snake animation


How I Work

Question → Hypothesis → Experiment → Measure → Audit → Improve
                         ↘ Build → Test → Document

I want my profile to show both sides of my work: research questions tested carefully and software projects built iteratively. I aim to make the evidence, limitations, and current status clear enough for collaborators, researchers, and recruiters to evaluate.


Connect

GitHub · LinkedIn · Portfolio

Open to internships, research conversations, collaborations, and applied AI engineering opportunities.

Built with curiosity · evaluated with care · improved through iteration.

Pinned Loading

  1. Bud-AI Bud-AI Public

    Stateful multi-agent AI tutor with memory, adaptive learning, and reward-driven educational benchmarking.

    Python 1

  2. Chat-AI Chat-AI Public

    AI-powered chat assistant with LLM integration, conversational AI, and a Streamlit interface.

    Python 1

  3. MCP MCP Public

    TypeScript 1

  4. forage-midas forage-midas Public

    Forked from vagabond-systems/forage-midas

    Project repo for the JPMC Advanced Software Engineering Forage program

    Java 1

  5. Ai-assisted-code-management-with-IBM Ai-assisted-code-management-with-IBM Public

    Python 1

  6. cineflex-recommendation-system cineflex-recommendation-system Public

    A sleek, high-performance movie and K-drama recommendation engine built with Next.js, Tailwind CSS, and TMDB API. Features dynamic algorithmic match scoring based on user interactions.

    TypeScript 1