Skip to content
View Achalnawal2745's full-sized avatar
☺️
Hello
☺️
Hello
2 followers · 1 following

Block or report Achalnawal2745

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. 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
Achalnawal2745/README.md

📌 Quick Nav

👋 About Me


"Building intelligent systems that solve real-world problems — not just prompting a model, but engineering the pipelines around it."

I'm a Computer Engineering student at Thapar Institute of Engineering and Technology, Patiala, India 🇮🇳, working in the space between Software Development Engineering (SDE) and production GenAI systems — asynchronous pipelines, agentic workflows, RAG architectures, and low-latency backends.

👉 Full case studies, live demos, and an interactive terminal walkthrough of everything below live on my portfolio: achalnawal2745.github.io/achal-portfolio

⚡ Quick Facts

  • 🔬 Past Experience: AI Applications Development Intern @ AmberFlux EdgeAI Pvt. Ltd. (May 2026 – Jul 2026)
  • 📄 Published: Co-author, Adv-HAMMER: Adversarially Robust Multimodal Fake Detection with Perturbation-Aware Training — accepted at PP-MisDet @ CVPR 2026
  • 🧑‍🏫 Open Source: Mentor, GirlScript Summer of Code (GSSoC) 2025
  • 🎓 Education: B.E. Computer Engineering, Thapar Institute of Engineering and Technology (CGPA 8.6/10)
  • 🟢 Status: Open for SDE / GenAI Systems roles

💼 Experience

🤖 AI Applications Development Intern | AmberFlux EdgeAI Pvt. Ltd.

May 2026 – Jul 2026 Python · FastAPI · Azure OpenAI · Azure Document Intelligence · Azure Service Bus · NetworkX · RapidFuzz

  • Built AI-powered document extraction workflows combining Azure Document Intelligence with GPT-based pipelines for automated engineering document processing.
  • Built parts of a concurrent, multi-threaded pipeline processing PDFs, emails, and technical drawings at scale, applying business-rule and data-quality checks.
  • Implemented FastAPI backend services and Azure Service Bus workflows for end-to-end pipeline orchestration.
  • Built a graph- and fuzzy-matching automation system (NetworkX + RapidFuzz) paired with LLM classification to cut down manual reconciliation in project/email-matching workflows.

🔬 AI/ML & Cybersecurity Research Intern | University of Queensland

Jul 2025 – Dec 2025 · via Thapar Institute of Engineering & Technology PyTorch · Adversarial ML · CNNs · Multimodal Detection

  • Implemented adversarial training for the HAMMER multimodal deepfake detection model, using a CNN-based generator to craft image perturbations.
  • Improved model robustness by training on mixed clean and adversarial samples, evaluated via AUC, EER, and IoU metrics.
  • Co-authored "Adv-HAMMER: Adversarially Robust Multimodal Fake Detection with Perturbation-Aware Training" — accepted at PP-MisDet @ CVPR 2026.

🚀 Featured Projects

🖥️ TITAN 2.0

Real-time multimodal desktop AI assistant

Python Gemini Live API WebSockets PyQt6 OpenCV Docling

Full-duplex voice assistant with sub-50ms audio streaming, biometric (face + voice) security, autonomous self-writing skill sandbox, and system-wide automation.

See more
  • Full-duplex voice interaction with GMM speaker verification and a system-wide instant-stop interrupt hook.
  • Bidirectional WebSocket bridge to a Chrome extension for browser automation with timeout recovery.
  • Autonomous skill-execution framework — TITAN writes, tests, and self-heals its own Python tools with AST-based safety checks.
  • Prompt-injection detection and sensitive-action sanitization guardrails.

GitHub ↗

🧩 ChatFusion

Multi-source RAG knowledge workspace

Python Flask RAG ChromaDB Sentence Transformers Whisper Gemini API

Conversational querying and synthesis across YouTube videos, PDFs, and audio — fully local embeddings + transcription for private, cost-free retrieval.

See more
  • Cross-document and cross-modal retrieval with synthesis over multiple sources at once.
  • Local-first: on-device embeddings (MiniLM) and transcription (faster-whisper); only retrieved chunks hit the LLM.
  • Flask REST backend with per-answer source attribution.

Live Demo ↗ · GitHub ↗

🛡️ PatchForge AI

4-agent autonomous DevSecOps pipeline

React TypeScript Vite Python Gemini 2.5 Flash

Chains Developer → CyberSec Auditor → Auto-Patcher → Git Automator agents to turn a plain-language prompt into a security-reviewed, PR-ready commit.

See more
  • Static-analysis auditor flags OWASP Top 10 / CWE vulnerabilities and auto-generates safe PoC exploits.
  • Maps findings to OWASP, SOC 2, GDPR, PCI-DSS and surfaces them in the final PR description.
  • Syntax-validation safety net requests a corrected fix from Gemini before anything is staged.

Live Demo ↗ · GitHub ↗

🌊 FloatChat

ARGO Ocean Intelligence System

Python FastAPI Streamlit PostgreSQL ChromaDB Gemini API NetCDF

Natural-language interface over global oceanographic float data — from raw NetCDF ingestion to conversational querying and dashboards.

See more
  • Automated pipeline to download, ingest, and store ARGO NetCDF files into PostgreSQL with SQL-based validation.
  • FastAPI backend paired with Streamlit dashboards for interactive exploration.

GitHub ↗

✋ Hand Gesture Drawing Pad

Touch-free real-time drawing app

Python OpenCV MediaPipe NumPy

Five-gesture recognition system (draw, erase, clear, resize brush, cursor mode) with dwell-click UI interaction — no mouse, no keyboard.

See more
  • Dwell-click system: hovering a UI button for 1s triggers a click with a live progress-ring indicator.
  • 20-stroke undo stack, 6 selectable colors, full keyboard-shortcut parity.

GitHub ↗

🌐 More on the Portfolio

Every project above has deeper technical write-ups, architecture notes, and an interactive terminal you can actually type commands into.

→ Explore the full portfolio

Try typing skills, projects, or whoami into the terminal on the site 👀

🚀 Tech Stack

💻 Languages


C++ Python JavaScript TypeScript SQL HTML5 CSS3

🧠 GenAI & Machine Learning


PyTorch TensorFlow Scikit-learn OpenCV MediaPipe YOLOv8 RAG Pipelines

☁️ Cloud & AI Platforms


Azure OpenAI Azure Document Intelligence Azure Service Bus Gemini API

⚙️ Backend & Databases


FastAPI Flask Node.js PostgreSQL MySQL ChromaDB

🛠️ Frameworks & Tools


React PyQt6 Docker Git GitHub WebSockets Streamlit Postman

📄 Publications & Certifications
📝 Adv-HAMMER: Adversarially Robust Multimodal Fake Detection with Perturbation-Aware Training PP-MisDet @ CVPR 2026 — View Paper ↗
🧑‍🏫 Open Source Mentor GirlScript Summer of Code (GSSoC) 2025
🎓 Machine Learning with Python Coursera
🎓 Data Structures and Algorithms in C++ Udemy
🎓 Introduction to Cybersecurity Cisco Networking Academy

📊 GitHub Analytics

GitHub Streak

🐍 Contribution Snake

github contribution grid snake animation

📈 Activity Pulse

Activity Graph

🔗 Connect with Me

Profile Views

"Code is like humor. When you have to explain it, it's bad."

Pinned Loading

  1. chatFusion chatFusion Public

    Python 2

  2. floatchat floatchat Public

    Python 2

  3. patchforge-ai patchforge-ai Public

    TypeScript 2 1

  4. drawpad drawpad Public

    Python 2

  5. orbical orbical Public

    JavaScript 2