Machine Learning Engineer
PayGlocal · Fintech Startup
Real-time payment risk systems. Designed time-series signal-based personalized risk model. Reduced false alerts by 50%+.
Time-SeriesRisk ModelingFintech
Ankush Gupta
He / They
Foundational & Multimodal AI
I am a PhD student in Computer Science at Purdue University (started Fall 2026), working with Dr. Aniket Bera in the IDEAS Lab. My research focuses on foundational multimodal AI, building generalizable models that integrate perception, language, and embodied reasoning for real-world robotic deployment.
Before Purdue, I completed my B.Tech in Computer Science at IIIT Delhi (2021–2025), where I worked on multimodal emotion understanding, and physiological signal modeling with Prof. Pushpendra Singh and Prof. Mohan Kumar, and on scholarly recommendation systems with Prof. Mukesh Mohania. This work led to publications at NeurIPS 2024 & 2025, ACM IMWUT 2025, and papers under review at TACL & EDBT.
I am interested in research collaborations and summer internships.
Academic Honors
- Started PhD at Purdue (Fall 2026)
- Published at NeurIPS '24 & '25
- oSTEM Grad Application Scholarship, 2025
- Google DeepMind Symposium Invitee, 2024
- NTSE Scholar, 2019
Research Interests
- Foundational Multimodal Models
- mbodied AI & Learning
- Perception & Language Grounding
- Human-Centered Intelligent Systems
Service & Community
- Workshop Chair, Queer in AI @ NAACL 2025
- Reviewer: ACL 2026, AAAI 2025, ACM CHI 2025, NeurIPS 2025
- Student Senate, IIIT Delhi
Teaching
- TA · Interactive Systems
- TA · Human-Centered AI
- TA · Intro to Programming
- TA · Operating Systems
Experience Timeline
An overview of my work across academia, research, and industry, highlighting key roles and contributions.
AI Undergrad Researcher · Data Science Lab
IIIT Delhi · Prof. Mukesh Mohania
Developed a taxonomy-aligned citation recommendation system combining hierarchical DAG representations, entity matching, and semantic retrieval for context-aware citation generation and placement.
Information RetrievalGraph-Based ModelingSemantic SearchNLP
HCI Researcher · Creative Interface Lab
IIIT Delhi · Dr. Anmol Srivastava
Gesture-tracking game for children using RAGs and RealSense depth motion. Augmented Hindi language learning. Published at ACM India HCI 2023.
HCIGesture TrackingAssistive Tech
B.Tech · Computer Science
IIIT Delhi
Research in ML, HCAI, HCI. TA for 4 courses. Student Senate & Disciplinary Action Committee.
AI ResearcherA* ConferenceStudent SenateTA × 4
PhD in Computer Science
Purdue University · IDEAS Lab · Dr. Aniket Bera
Researching foundational multimodal AI, building generalizable models at the intersection of perception, language, and embodied intelligence for real-world robotic deployment.
Foundational Multimodal Models
Embodied AI
Robotic Deployment
Perception & Language
HCAI Research Assistant · Melange Lab
IIIT Delhi · Prof. Pushpendra Singh & Prof. Mohan Kumar (RIT)
Built multimodal emotion recognition systems using physiological signals, spanning large-scale benchmarking, dataset development, and real-world data collection frameworks.
Physiological SignalsMultimodal LearningAffective ComputingBenchmarking
MLE Intern
E.ON Digital Technology
Developed a CI/CD pipeline chatbot for power grid operations, reducing workflow overhead and improving usability, alongside optimizing ML/NLP models for intelligent automation.
MLOpsCI/CD SystemsConversational AIIndustrial AI
Student Summer Intern
E.ON SE
Automated Microsoft Azure pipeline. Increased image classification efficiency to 95%, reduced computational overhead by 5%.
AzureComputer VisionMLOps
Selected Research
Published in core CS venues. Filter by theme below. Full list on Google Scholar
NeurIPS 2025
FEEL: Quantifying Heterogeneity in Physiological Signals for Generalizable Emotion Recognition
A large-scale benchmarking framework for emotion recognition from physiological signals, unifying evaluation across 19 EDA and PPG datasets. Benchmarks 16 architectures under within- and cross-dataset settings, revealing strong generalization of contrastive signal-language pretraining.
NeurIPS 2024
EEVR: A Dataset of Paired Physiological Signals and Textual Descriptions for Joint Emotion Representation Learning
A multimodal dataset pairing EDA and PPG signals from 360° VR-based emotion elicitation with participant-level textual descriptions (37 subjects). Introduces contrastive language-signal pretraining (CLSP), enabling strong zero-shot transfer across datasets.
ACM IMWUT 2025 · UbiComp
AnnoSense: Navigating the Complexities of Collecting and Annotating Everyday Emotion Data for AI
A stakeholder-informed framework for real-world emotion data collection, developed through studies with 119 participants. Produces 15 actionable guidelines for reliable, usable, and ethically grounded emotion annotation.
Under Review
CITAR: Citation Inference using Taxonomy-Aligned Retrieval and Context-Aware Entity Matching
A context-aware citation recommendation system aligning hierarchical academic taxonomies. Achieves strong retrieval performance (Recall@10 0.35, MRR 0.21) while improving exposure balance across the corpus.
Under Review
Toward Inclusive NLP: A Survey of LGBTQIA+ Representation and Gaps in Language Technologies
A systematic survey of 86 NLP papers examining language technologies and LGBTQIA+ communities. Identifies key gaps in bias mitigation, intersectional representation, and non-English settings.
ACM India HCI 2023 · Springer SCI
Jail Diaries: An Interactive Pop-Up Book Art Installation
Interactive learning experiences for children about the Indian freedom struggle, using visual projection mapping and gesture tracking. Published in Springer "Studies in Computational Intelligence."
News & Activity
Recent highlights. LinkedIn ↗
Fall 2026Education category
Joining PhD in Computer Science at Purdue University
Excited to begin my PhD under Dr. Aniket Bera at the IDEAS Lab, working on foundational multimodal AI for real-world robotic deployment.
→ LinkedIn Post
NeurIPS 2025Conference category
FEEL Paper Accepted at NeurIPS 2025
Benchmarking framework for physiological signal emotion recognition across 19 datasets accepted at NeurIPS 2025 D&B track.
→ Paper Link
NAACL 2025Conference category
Workshop Chair at Queer in AI @ NAACL 2025
Archival workshop chair for the Queer in AI workshop — organizing a space for LGBTQ+ researchers at NAACL 2025.
→ Proceedings
2024Visiting category
Invited to Google DeepMind Research Symposium
Selected as invited attendee at the Google DeepMind Research Symposium 2024.
→ LinkedIn PostThe Other Side
Research is only one part of who I am. Outside the lab, I explore ideas through writing, debate, and community.
Poetry & Writing
I write poetry as a way of thinking and exploring about identity, belonging, and the quiet tensions of society. Click to read.
Debating
Competitive debating sharpened how I think about arguments, evidence, and persuasion, gives safe space for putting up my political views.
Reading
Mostly rom-com novels these days, with a soft spot for queer stories — light, funny, and easy to keep returning to.
Currently reading — click to view
Student Governance
Student Senate member at IIIT Delhi, contributed to institutional policy and student representation, while mentoring junior students and supporting their academic journey.
Queer in AI
Active Core organizer and NAACL 2025 Workshop Chair at Queer in AI, building community and supporting visibility for LGBTQ+ researchers in AI/ML.
Teaching
TA'd four courses including Human-Centered AI and Interactive Systems. Good teaching is good research communication — both require knowing what your audience actually needs.
Let's Connect
I'm actively looking for research collaborations and industrial internship opportunities (summer 2027), particularly at the intersection of multimodal AI and applied machine learning.
Whether you're a researcher interested in joint work on foundational models, or an industry team looking for a collaborative PhD intern, I'd love to hear from you.
For academic inquiries, reviews, or just to talk about ideas or collaborations, feel free to reach out.