Mohib Ali Khan
Build Scalable Web Apps That Turn Data Into Decisions
New to Contra
Mohib Ali is building their profile!
๐จ 92% accuracy sounds impressive until you see what it misses.
Excited to share that our research paper has been published in the International Journal of Innovations in Science and Technology
๐๐ฒ๐๐ฒ๐ฟ๐ฎ๐ด๐ถ๐ป๐ด ๐๐ฒ๐ฒ๐ฝ ๐๐ฒ๐ฎ๐ฟ๐ป๐ถ๐ป๐ด ๐ฎ๐ป๐ฑ ๐ก๐๐ฃ ๐ณ๐ผ๐ฟ ๐๐ต๐ฒ ๐๐ฑ๐ฒ๐ป๐๐ถ๐ณ๐ถ๐ฐ๐ฎ๐๐ถ๐ผ๐ป ๐ผ๐ณ ๐๐ฒ๐ฐ๐ฒ๐ฝ๐๐ถ๐๐ฒ ๐ข๐ป๐น๐ถ๐ป๐ฒ ๐๐ผ๐ป๐๐ฒ๐ป๐
We developed and compared multiple ML and NLP models including Logistic Regression, SVM, Bi-LSTM, and BERT for detecting deceptive online content.
Key result:
BERT significantly outperformed traditional models, achieving 92.5% accuracy by better capturing contextual meaning in text.
๐ก But what we learned goes beyond numbers:
โ High accuracy does not always mean real-world reliability.
โ Explainability is still a major gap in NLP systems.
โ Fake news detection is an evolving and continuous problem.
This work made one thing clear building a model is easy, but building a reliable system for real-world misinformation is much harder.
Special thanks to my mentors Faheem Ahmed, Ph.D (https://www.linkedin.com/in/faheem-ahmed-ph-d-4a057723/) | Kamran Dahri (https://www.linkedin.com/in/kamran-dahri-70a06925a/) | Muhammad Aquib (https://www.linkedin.com/in/muhammad-aquib-a31a49192/) | Dr Muhammad Yaqoob Koondhar (https://www.linkedin.com/in/dr-muhammad-yaqoob-koondhar/) for their guidance and continuous feedback throughout this journey.
๐ full paper: https://journal.50sea.com/index.php/IJIST/article/view/1834
Our next goal is to publish in higher-tier, indexed journals and contribute to research that genuinely impacts how we fight misinformation online.
Do you think AI alone can solve misinformation, or will humans always be part of the solution?
#DataScience (https://www.linkedin.com/search/results/all/?keywords=%23datascience&origin=HASH_TAG_FROM_FEED) #MachineLearning (https://www.linkedin.com/search/results/all/?keywords=%23machinelearning&origin=HASH_TAG_FROM_FEED) #NLP (https://www.linkedin.com/search/results/all/?keywords=%23nlp&origin=HASH_TAG_FROM_FEED) #DeepLearning (https://www.linkedin.com/search/results/all/?keywords=%23deeplearning&origin=HASH_TAG_FROM_FEED) #BERT (https://www.linkedin.com/search/results/all/?keywords=%23bert&origin=HASH_TAG_FROM_FEED) #Research (https://www.linkedin.com/search/results/all/?keywords=%23research&origin=HASH_TAG_FROM_FEED) #SiliconValley (https://www.linkedin.com/search/results/all/?keywords=%23siliconvalley&origin=HASH_TAG_FROM_FEED) #TechCommunity (https://www.linkedin.com/search/results/all/?keywords=%23techcommunity&origin=HASH_TAG_FROM_FEED)
0
0
๐๐ฒ๐น๐ถ๐๐ฒ๐ฟ๐ฒ๐ฑ ๐๐ป๐ฑ-๐๐ผ-๐๐ป๐ฑ ๐ค๐ ๐ณ๐ผ๐ฟ ๐ฆ๐๐ข & ๐๐ ๐ฆ ๐ ๐ถ๐น๐ฒ๐๐๐ผ๐ป๐ฒ (๐จ๐-๐๐ฎ๐๐ฒ๐ฑ ๐ฃ๐น๐ฎ๐๐ณ๐ผ๐ฟ๐บ)
Executed structured testing for a live talent acquisition platform, ensuring smooth rollout of SEO enhancements without impacting existing user flows.
What was covered:
SEO landing pages logic & visibility controls
CMS-driven dynamic content (hero sections, highlights, banners)
Admin vs frontend data consistency
UI rendering validation across key pages
Regression assurance across:
Authentication system (Login / Register)
Real-time messaging
Job creation & management
Talent discovery (search & filters)
Execution approach:
Tested 40+ scenarios across multiple modules.
Followed structured acceptance + regression checklist
Focused on real user journeys, not just isolated test cases
Key learning:
Effective testing isnโt just about finding bugs
itโs about ensuring new changes donโt break existing user flows.
Outcome: Smooth deployment, stable system, zero critical breakage.
2
2
15