Skip to content
View aaqibnp971's full-sized avatar
  • 18:15 (UTC +04:00)

Highlights

  • Pro

Block or report aaqibnp971

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
aaqibnp971/README.md

Aaqib Pangarkar

Software Developer • Machine Learning • AI Pipelines & Automation

Building end-to-end AI systems, data pipelines, and automation workflows — from research to production.


About

  • Computer Science Engineering student at BITS Pilani Dubai Campus (graduating July 2027)
  • Software Developer at R13 Labs — building the audio AI engine powering a psychoacoustic wellness platform
  • Background in NLP, signal processing, RAG pipelines, wearable biosignal systems, and workflow automation
  • Open to internship opportunities in ML engineering, AI systems, and software development

Core Projects

Audio Stem Splitter — Production-Grade Music Source Separation

6-stem audio separation pipeline powering R13 Labs' Prism psychoacoustic engine.

  • Meta's HTDemucs model achieving ~8–9 dB SDR on vocals across 6 isolated stems
  • ffmpeg preprocessing pipeline handling format normalization, silence trimming, and peak normalization
  • Automatic CUDA/MPS/CPU fallback for optimized inference across compute environments
  • Browser-based web UI with live per-stem audio playback and individual WAV downloads

Hallucination Detection Pipeline — NLP Research

Training-free, token-level hallucination detection framework benchmarked against state-of-the-art methods.

  • KL divergence, information gain, and confidence drop signals derived from dual GPT-2 forward passes
  • 0.7243 AUROC on RAGTruth — outperforming SelfCheckGPT (0.6805)
  • 0.9249 AUROC on HaluEval with zero fine-tuning or labelled training data
  • Benchmarked across 6 LLMs including Llama-2-70B, GPT-4, and Mistral-7B

Wearable Stress Detection & Real-Time Sound Prescription

Biosignal-based stress classification with a real-time audio intervention layer.

  • Random Forest on WESAD and DEAP datasets — 91.33% accuracy, 0.940 F1-score
  • 10 physiological features engineered from EDA, BVP, and skin temperature signals
  • Cross-dataset emotion zone mapping bridging stress labels to valence-arousal quadrants
  • Streamlit app dynamically generating personalised alpha binaural beat audio from live sensor input

Blockchain Academic Credential Verification

Decentralised credential system for secure university degree issuance and employer verification.

  • Hyperledger Fabric with Go smart contracts for on-chain issuance, tracking, and revocation
  • Node.js SIS bridge generating W3C Verifiable Credential JSON for secure issuance
  • Streamlit interfaces for university and employer users to manage and verify credentials

Technical Stack

Languages

Python Java C

ML / AI

PyTorch scikit-learn HuggingFace

Data & Automation

n8n MySQL Supabase Pinecone

Frontend & Deployment

Streamlit Tailwind CSS HTML/CSS/JS


Experience

Software Developer — R13 Labs India Pvt Ltd (May 2026 – Present)

  • Engineered the audio source separation pipeline (HTDemucs) powering R13 Labs' Prism psychoacoustic engine
  • Built audio preprocessing, GPU acceleration fallback, and a full browser-based playback interface
  • Part of the core team awarded Innovative Education Platform of the Year — CorporateLiveWire 2026 (UK)

Software Intern — TechnoCIT Software Solutions, Dubai (June – August 2025)

  • Built an agentic AI banking chatbot and multi-country web scraping system using n8n
  • Designed RAG pipelines with vector embeddings, semantic search, and metadata-tagged storage in Supabase and Pinecone
  • Built fault-tolerant workflows with retry logic, exponential backoff, structured logging, and alerting

Achievements

  • 🏆 Innovative Education Platform of the Year — CorporateLiveWire Innovation & Excellence Awards 2026 (R13 Labs)

Currently

  • Studying NLP, Computer Networks, and Algorithms at BITS Pilani Dubai
  • Building psychoacoustic AI systems and audio pipelines at R13 Labs
  • Exploring advanced RAG architectures and production ML deployment patterns

Connect

📧 aaqibnp@gmail.com
🔗 LinkedIn
🐙 GitHub

Popular repositories Loading

  1. NLP_Assignment NLP_Assignment Public

    An end-to-end Machine Learning pipeline to detect and classify AI hallucinations in Natural Language Processing, leveraging the HaluEval and RAGTruth datasets.

    Jupyter Notebook 1

  2. DATN-Blockchain-Project DATN-Blockchain-Project Public

    A decentralized credentialing system using Hyperledger Fabric and W3C Verifiable Credentials to securely issue, store, and verify university degrees.

    Python

  3. FDS_Project FDS_Project Public

    Stress detection from wrist signals with 91.33% accuracy. Random Forest on WESAD + DEAP datasets with real-time binaural beat intervention. Python, Scikit-learn, Streamlit.

    Python

  4. audio-stem-splitter audio-stem-splitter Public

    Production-grade audio stem splitter powered by Meta's HTDemucs — splits any track into 6 isolated stems (vocals, drums, bass, guitar, piano, other) via a web UI and REST API. Built for the Prism a…

    Python 1

  5. aaqibnp971 aaqibnp971 Public