Manikandan Ravikiran is an influential (to the select few around him, and occasionally to Reviewer 2) AI researcher/engineer whose work has wandered across machine learning, large language models, multilingual and educational NLP, trustworthy AI, reasoning systems, efficient learning, computer vision, and several research problems that, in retrospect, probably should have been left unsupervised.

Over the years, his work has included AI for Advanced Driver Assistance Systems (ADAS), scalable machine-learning systems, model compression, multilingual NLP, educational AI, LLM evaluation, semantic completeness, explanation quality and fairness, and mathematically grounded approaches to efficient learning and reasoning. Some of these became papers. Some became patents. Some became rebuttals. A concerning number became all three.

He has submitted his doctoral thesis in Computer Science and Engineering at IIT Mandi, advised by Prof. Rohit Saluja and Prof. Arnav Bhavsar. The thesis is finished; the defense is pending and is currently operating on the well-established academic principle that time is merely a suggestion.

After spending an unreasonable number of years thinking about language models — and repeatedly participating in the recurring social experiment otherwise known as ACL/ARR reviewing — Manikandan is currently taking a long hiatus from Research, LLM and generative-AI research to investigate a considerably harder open problem:

whether life contains anything beyond leaderboards, ablation tables, rebuttal deadlines, reviewer confidence scores, and GitHub repositories called final_final_v7_fixed.

Preliminary results are inconclusive.

For the foreseeable future, he is not pursuing new LLM collaborations, research engagements, benchmarking exercises, consulting projects, advisory roles, “quick experiments,” brainstorming calls, unpaid technical archaeology, or meetings that could have been emails.

Messages containing any of the following expressions may therefore undergo aggressive pruning: LLM · RAG · Agentic AI · Prompt Engineering · Evaluation Framework · Benchmark · SOTA · Foundation Model · “Can we pick your brain?” · “Just 20 minutes” · “We already have the idea”

The last phrase is particularly dangerous when followed by:

“…we just need someone to implement it.”

Such requests are automatically routed through a proprietary rejection architecture featuring zero-shot filtering, adversarial regularization, and an exceptionally low recall objective.

This website is therefore best understood as an archive of old research, publications, projects, patents, academic adventures, experiments, mathematical detours, and ideas that somehow survived peer review long enough to acquire DOIs.

Some pages are current.
Some are historical.
Some probably need updating.
Some contradict newer pages.
And some continue to exist solely because deleting things from a personal website requires more motivation than writing another paper.

Contact

LinkedIn is dead.
Twitter is still being mourned in its previous incarnation.
Email is currently out with his girlfriend and appears to be getting serious.

Consequently, contacting Manikandan may require one of the following:

1. remarkable persistence,
2. improbable luck,
3. a genuinely interesting (to him) problem,
4. physically encountering him somewhere (in foothills of himalayas),
5. or successfully demonstrating that your proposed “15-minute chat” will, in accordance with the laws of physics, actually last 15 minutes.

If you do manage to reach him, please use the opportunity responsibly.

He has already seen your benchmark.

🔔 Selected Updates

• 2026 — Submitted Ph.D. thesis in Computer Science and Engineering at IIT Mandi; defense pending.
• April 17, 2025 — Delivered an invited guest lecture on Multimodal Generative AI and Theoretical Diffusion Modeling at VIT Chennai.
• March 20, 2025 — Congratulations to Hariharan (NIT-K) on the successful submission of his Ph.D. thesis.
• March 17, 2025 — Delivered an invited lecture on Advanced Reinforcement Learning at NIT Surathkal — thank you to the Department of IT, NIT-K.
• Nov 20, 2024 — Delivered a keynote on Industrial Computer Vision at IEEE ICDDS 2025.
• Nov 10, 2024 — Three papers accepted at IEEE BigData 2024 — congratulations to all co-authors!
• Dec 15, 2024 — Successfully organized the workshop on Handling Resource Constraints using Big Data and AI.
• Dec 10, 2024 — Presented ongoing work at the LC4 Workshop on Real World Applications as part of SPELLL.
• Jan 10, 2023 — Honored to receive the ACL Best Reviewer Prize 2023 — thank you ACL!
• Sep 10, 2023 — Delivered a guest lecture on Multilingual NLP at VIT Chennai.
• 2022–Present — Mentoring researchers and Ph.D. scholars on multilingual NLP, educational AI, and related machine-learning problems.

🤝 Collaborators

Over the years, I have had the opportunity to collaborate with researchers from several universities and industrial research laboratories, including:

• Dr. Martin Klignkit — Kyocera Innovation Labs, Japan
• Dr. Shinichi Satoh and team — National Institute of Informatics, Japan
• Dr. Andreas Dengel, Sheraz Ahmed, Jorn Hees — DFKI, Germany
• Dr. Snehanshu Saha — APP CAIR, BITS Pilani
• Dr. Vinay Namboodiri — DelTA Lab, IIT Kanpur
• Dr. Bharathi Raja Asoka Chakravarthi — National University of Ireland, Galway
• Dr. Gaurav Sharma — IIT-K / INRIA-THOTH / NEC Media Analytics USA
• Dr. Anand Kumar Madasamy — NIT Karnataka, Surathkal
• Dr. Sangeetha Sivanesan — NIT Tiruchirapalli
• Dr. Ratnavel Rajalakshmi — VIT Vellore

📚 Selected Publications

Educational AI and NLP

• GISA: Gradual Information Selection Attention for MCQ Difficulty Estimation
  Manikandan Ravikiran, Tarun Sharma, Rajat Verma, Rohit Saluja, Arnav Bhavsar
  AIED 2025 (CORE A) — Accepted

• TEEMIL: Towards Educational MCQ Difficulty Estimation in Indic Languages
  Manikandan Ravikiran, Siddharth Vohra, Rajat Verma, Rohit Saluja, Arnav Bhavsar
  COLING 2025

Machine Learning and Computer Vision

• AEI-DRL: Adaptive Ensemble Imputation for Trip Data using Deep Reinforcement Learning
  Ankit Sharma, Akhash Vellandurai, Thiruvengadam Samon, Vinoth Kumar, Manikandan Ravikiran
  IEEE BigData 2024

• DKT: A First Look at Dynamic Kernel Tuning for Pedestrian Attribute Recognition
  Manikandan Ravikiran, Rahul Mishra, Soumen Biswas, Ananth Ganesh
  *IEEE BigData 2024
  Equal contribution

• Hi-GOTE: Hierarchical Groupwise Temporal Ensembling for Pedestrian Attribute Recognition
  Manikandan Ravikiran, Soumen Biswas, Ananth Ganesh
  IEEE ICMLA 2023

• You Reap What You Sow: Revisiting Intra-Class Variations and Seed Selection in Temporal Ensembling
  Manikandan Ravikiran, Siddarth Vohra, Yuichi Nonaka, et al.
  COMSYS 2021

• A Sensitivity Analysis (and Practitioners’ Guide to) of DeepSORT for Low Frame Rate Video
  M. Ravikiran, Y. Nonaka, N. Mariyasagayam
  IEEE Big Data 2020

• Multilayer Pruning Framework for Compressing Single Shot Multibox Detector
  Pravendra Singh, Manikandan Ravikiran, Neeraj Matyali, Vinay P. Namboodiri
  IEEE WACV 2019

Multilingual and Low-Resource NLP

• Revisiting Automatic Speech Recognition for Tamil and Hindi Connected Number Recognition
  Rahul Mishra, Senthil Raja Gunaseela Boopathy, Manikandan Ravikiran, et al.
  Third Workshop on Speech and Language Technologies for Dravidian Languages, 2023

• Findings of the Second Shared Task on Offensive Span Identification in Code-Mixed Tamil-English Comments
  Manikandan Ravikiran, Bharathi Raja Chakravarthi, et al.
  Third Workshop on Speech and Language Technologies for Dravidian Languages, 2023

• MMOD-MEME: A Dataset for Multimodal Face Emotion Recognition on Code-Mixed Tamil Memes
  Ramesh Kannan, Manikandan Ravikiran, Ratnavel Rajalakshmi
  LC4 Workshop, CCIS Springer Series, 2023

• Overlapping Word Removal is All You Need: Revisiting Data Imbalance in Hope Speech Detection
  Hariharan RamakrishnaIyer LekshmiAmmal, Manikandan Ravikiran, et al.
  Journal of Experimental and Theoretical Artificial Intelligence, Taylor & Francis, 2023

• DOSA: Dravidian Code-Mixed Offensive Span Identification Dataset
  Manikandan Ravikiran, Subbiah Annamalai
  Workshop on Dravidian Languages, EACL 2021

🧪 Selected Research Projects

Some research directions explored over the years include:

• Latent Plan Execution for Text Completeness
  Graph-based approaches for studying intent alignment and semantic completeness in generated text.

• CASSA / GISA Attention
  Attention mechanisms for modeling MCQ difficulty using inductive reasoning, psychometric modeling, and analytical geometry.

• PEARL for Explanation Auditing
  Methods for studying explanation sufficiency and reasoning traceability in machine-learning systems.

• Dialectual Educational Platforms
  Culturally and linguistically adaptive learning systems for underrepresented languages and dialects.

These projects are listed for archival purposes. No inference should be made from their presence here regarding current research activity. The robots are resting.

🎓 Teaching and Professional Service

Professional Service

TPC Member

• ACM ICMR — 2019
• IEEE ICDDS — 2022
• FIRE — 2021, 2022
• DravidianLangTech — 2021, 2022, 2023, 2024
• LTEDI — 2021, 2022, 2023, 2024

Reviewer

• ACL — 2017, 2020, 2023, 2024, 2025
• EMNLP — 2022, 2023, 2024
• AACL — 2020, 2022, 2024
• EACL — 2021, 2023
• AIED — 2025
• NAACL — 2018, 2024
• COLING — 2018, 2024
• IEEE BigData — 2020
• ACM ICMR — 2019
• IEEE ICMLA — 2023
• IEEE ICDDS — 2022
• FIRE — 2021, 2022
• DravidianLangTech — 2021, 2022, 2023
• LTEDI — 2021, 2022, 2023
• Springer Language Resources and Evaluation
• Elsevier Engineering Applications of Artificial Intelligence
• ACM Transactions on Asian and Low-Resource Language Information Processing
• Springer Nature Computer Science
• Taylor & Francis Journal of Experimental and Theoretical Artificial Intelligence

Organizer

• Workshop on Handling Resource Constraints for/using ML — IEEE BigData 2024
• Fourth Workshop on Speech and Language Technologies for Dravidian Languages — EACL 2024
• Special Session on Machine Learning for Graphs — IEEE ICMLA 2023
• Special Session on Handling Resource Constraints for/using ML — IEEE ICMLA 2023
• Workshop on Low Resource Cross-Domain, Cross-Lingual & Cross-Modal Offensive Content Analysis — SPELLL 2022, 2023
• Workshop on Cross Modal Learning and Application — ACM ICMR 2019

Other Roles

• Mentor — ACM CSCW 2019 Student Reviewer Mentor
• Publicity Chair — SPELLL 2022
• Session Chair — DravidianLangTech
• Industry Session Chair — ACM ICMR

💡 Patents

• System and Method Inference Scaling with Weight Reusability
  Manikandan Ravikiran, Ananth Ganesh — 2024, Under Review

• System and Method for Finding Novel Objects Across Domains via Weight Switching and Tracing
  Manikandan Ravikiran, Ananth Ganesh, Yuichi Nonaka — 2022, Granted

• System and Method for Generalization through Reimann Conditioned Representation
  Manikandan Ravikiran, Yuichi Nonaka, Kingshuk Banerjee — 2021, Under Review

• System and Method to Generate Gating Sequences for Training Model on Multidomain Datasets
  Shibashish Sen, Manikandan Ravikiran, Yuichi Nonaka, Nestor Mariyasagayam — 2021, Granted

• System and Method for Generating Filter Sequences to Train Model on Completely Noisy Dataset
  Manikandan Ravikiran, Yuichi Nonaka, Nestor Mariyasagayam — 2021, Granted

• System and Method to Train Neural Network with Heterogenous Distillation
  Manikandan Ravikiran, Shibashish Sen — 2020, Granted

• System and Method to Train Object Recognition Network with Less Data
  Manikandan Ravikiran — 2019, Granted

• Method and System for Generating an Optimal Object Recognition Network (OORN) with Balanced Accuracy
  Manikandan Ravikiran — 2017, Granted

• Method and System for Determining Plausibility of a Clinical Care Plan
  Manikandan Ravikiran, Sarath P. R., Saima Mohan — 2015, Granted

🔗 Links

• GitHub
• 📫 Email: manikandan.ravikiran@gmail.com

Google Scholar is currently taking a sabbatical too.