INFERENCELAB
Applied AI Lab · Multan, Punjab, Pakistan

Research. Engineering. Education. Built to ship.

INFERENCE Lab conducts reproducible AI research, builds production-grade systems, and trains engineers who can design, deploy, and scale AI in the real world. Evidence over hype. Engineering over branding.

Largest
ROMAN URDU LANGUAGE RESOURCE
SOTA
MODELS FOR LOW-RESOURCE LANGUAGE & HEALTH RESEARCH
GLOBAL
RESEARCH COLLABORATIONS & PUBLICATIONS
4+
OPEN-SOURCE AI LIBRARIES
What we are

One organization, three reinforcing tracks.

INFERENCE Lab is an applied AI research and engineering organization founded by Muhammad Khubaib Ahmad. It operates as original research, AI/software engineering services, and structured engineering education — under one identity, each track reinforcing the others' credibility.

Our aim: close the gap between people who know AI concepts and engineers who can build, deploy, and maintain AI systems.

Engineering discipline over hype

Real, deployed output — not demos that break the moment they leave a notebook.

Evidence over branding

Reproducible pipelines, proper evaluation, and a permanent DOI on every research release.

Output over certificates

You leave with systems on GitHub and models on HuggingFace, not a PDF.

What we do

Three tracks, one engineering standard.

Research informs the curriculum. The curriculum trains the engineers. The engineers ship the services. Each track makes the others credible.

01

Research

Original research in low-resource NLP and speech intelligence — conducted independently and through international academic collaboration.

  • King Saud University · EPU Kuwait · Doane University (USA) · Hanyang University (Korea)
  • Every release ships reproducible pipelines + a permanent DOI
  • Deployable inference code, not leaderboard numbers
02

Engineering Services

AI/ML engineering, LLM and RAG systems, computer vision, speech AI, and full-stack delivery for organizations.

  • Production REST APIs, containerized & monitored
  • Same engineering discipline taught in the lab
  • Owned end-to-end across the full ML lifecycle
03

Engineering Education

A structured, deployment-focused curriculum that takes developers from basic Python syntax to deploying production AI systems.

  • 12.5 months · 6 phases · live coding every week
  • Engineering Fellowship for project-based contribution
  • Built, deployed, and on GitHub — not certificates
Research output

Low-resource NLP & speech intelligence.

Independent and international research — every release ships reproducible pipelines, evaluation documentation, and a permanent DOI.

  • Accepted2026

    Automated Vocal Fatigue Screening in Professional Voice Users: Development and Occupational Validation of an Automated Assessment System

    Development and occupational validation of an automated vocal load assessment tool for professional voice users — clinical-grade speech analysis in production.

    Journal of Voice · Hanyang University, Republic of Korea
  • Under Review2026

    Disagreement-Weighted Classification and Calibration-Aware Training for Subjective NLP Tasks: A Multi-Seed Empirical Study

    Introduces a disagreement-weighted classification framework that explicitly models annotator disagreement to improve learning in subjective NLP task.

    Language Resources and Evaluation · National University of Singapore
  • Under Review2026

    Ergonomic Intervention as Joint Cognitive System Reconfiguration: A Qualitative Case Study in Healthcare

    Explores ergonomic intervention as a joint cognitive system reconfiguration to improve coordination between healthcare professionals, technologies, and work environments.

    KU Leuven (Belgium), King Saud University
  • Published Preprint2026

    RUEmoCorp: A Large-Scale Roman Urdu Emotion Corpus & Benchmark Suite

    First large-scale Roman Urdu emotion corpus — 134K labeled samples with Fleiss κ = 0.658 (substantial agreement), multi-institute annotation, fully open-source on HuggingFace and Harvard Dataverse.

    Language Resources and Evaluation (Springer)DOI
  • Published Preprint2025

    RUDaSA: Roman Urdu Dataset for Sentiment Analysis — A Large-Scale, Curated Corpus with Privacy-Preserving Embeddings and Competitive Benchmarking of Transformer Models

    Large-scale Roman Urdu sentiment corpus built via privacy-preserving embedding pipelines. Benchmarks state-of-the-art Transformer models — addressing a critical gap in low-resource South Asian NLP.

    Research Square · PreprintDOI
  • Published Preprint2025

    Data-Centric Roman Urdu NLP: Dataset Curation & Model Benchmarking

    Largest high-quality Roman Urdu sentiment dataset via privacy-preserving embedding pipelines — SOTA 0.84 accuracy, 0.83 Macro-F1.

    Zenodo · PreprintDOI
  • Published Preprint2025

    Forecast-Based Decision Support System for Mango Malformation

    Time-series forecasting and smart-agriculture DSS — demonstrated 50–60% yield improvement through data-driven intervention.

    Zenodo · PreprintDOI
  • In Progress2026

    DASER-Net: Disentangled Adversarial Speech Emotion Recognition with Hierarchical Bottleneck Fusion for Cross-Corpus Generalization

    Cross-corpus speech emotion recognition study developing disentangled and domain-invariant representations for robust emotion classification across unseen acoustic domains.

    Open to collaboration
  • In Progress2026

    Cursor Kinematics as a Privacy-Preserving Indicator of Cognitive Fatigue and Stress in Knowledge Work: An Explainability-Driven Analysis of Within-Person Interaction Logs

    Within-person study investigating cursor kinematics as a privacy-preserving indicator of cognitive fatigue and acute stress using explainable machine learning on workplace interaction logs.

    Open to collaboration
  • In Progress2026

    A Post-Quantum Hybrid Key Encapsulation Mechanism with Lorenz Chaotic Diffusion for Authenticated Medical Image Encryption

    Combines post-quantum cryptography and chaos-based diffusion to provide strong confidentiality, integrity, and resistance against quantum-enabled attacks.

    Open to collaboration
Software & systems released

Open source, deployed, and used by researchers.

Libraries, APIs and platforms spanning speech AI, LLM observability, RAG, agents and MLOps infrastructure.

Speech AI · Library

auralis-vfs

Pip-installable Python library for vocal fatigue scoring from raw speech. Built on ECAPA-TDNN-VHE — returns a normalized fatigue score from any audio input.

PyPIECAPA-TDNNSpeech
Speech AI · Real-Time

voiceMonitor

Real-time vocal health monitoring tool — streams microphone input, runs auralis-vfs, and surfaces fatigue level and vocal load over a session.

Real-TimeMonitoringCLI
Speech AI · Verification

VocalID

Open-source speaker verification framework using embedding cosine similarity on ECAPA-TDNN representations. Lightweight, no external APIs required.

Speaker VerificationEmbeddingsOpen Source
Open Source · Library

faker-pk

Pakistani synthetic data library for realistic test fixtures — Urdu and Roman Urdu names, CNIC numbers, phone formats, addresses, and locale-aware records.

PyPISynthetic DataPakistan
LLM Observability

QueryVault

End-to-end LLM platform with real-time hallucination detection, structured logging, and a developer dashboard — hallucination rate, latency, P50/P95.

StreamlitFew-ShotObservability
Multi-Agent System

DataForge

Multi-agent system for automated EDA, cleaning, visualization, and report generation — autonomous data analysis pipelines built on CrewAI.

CrewAIEDAAgents
Image Encryption

SecureCipher v2.0

Image encryption scheme scoring 100/100 against 14 benchmark algorithms across standard cryptographic security tests.

CryptographyBenchmarked
Engineering education

From Python syntax to deployed AI systems.

A structured, deployment-focused curriculum across 6 phases and 12.5 months. Every 1.5-hour session is live coding — Saturday introduces the concept, Sunday goes deeper into edge cases, and assignments ship to GitHub by Friday.

Full curriculum
  1. Phase 02 months

    Engineering Foundations

    From a working terminal to production database architecture. Writing Python that does not break.

    8 weeks

  2. Phase 12 months

    Data Engineering & Visualization

    NumPy through publication-quality visualization, grounded in real statistical thinking.

    8 weeks

  3. Phase 22.5 months

    Machine Learning Engineering

    Classical ML the engineering way: pipelines, leakage-free evaluation, and experiment tracking.

    7 weeks

  4. Phase 32.5 months

    Deep Learning & NLP

    Neural networks from scratch to fine-tuned Transformers and speech AI pipelines.

    8 weeks

  5. Phase 42 months

    AI Systems & LLM Engineering

    Production AI APIs, disciplined LLM engineering, RAG systems, and agents that actually work.

    4 weeks

  6. Phase 51.5 months

    MLOps & Deployment

    Containerize, automate, ship, and monitor. The difference between a notebook and a product.

    3 weeks

Join INFERENCE Lab

Work on real AI systems, research, and open-source.

From published research with public DOIs to open-source libraries used by other engineers, INFERENCE Lab's work is real and verifiable. We build production AI systems, conduct original research in low-resource NLP and speech intelligence, and release our code, models, and datasets in the open — not behind a paywall or a portfolio screenshot.

Work with the lab

Train as an engineer, collaborate on research, or build a system with us.

Currently running our first online cohort and preparing the first Engineering Fellowship. Reach out about training, research collaboration, or AI engineering services.