Hi! I am Het, a recent Computer Science graduate from IIIT Delhi. I have experience working on projects around Natural Language Processing (NLP), Deep Learning (DL), and Reinforcement Learning (RL). I have done various internships and published research papers at top venues like ACL and AAAI. I have build and deployed machine learning systems and agentic workflows.
I am open to discussing new job and research opportunities around ML, NLP, DL, and RL. Feel free to reach out at het.shah25052004@gmail.com!
My research interests span Natural Language Processing, Large Language Models, Knowledge Graphs, Agentic AI, Retrieval-Augmented Generation, Deep Learning, Generative AI, and Reinforcement Learning.
- In NLP/LLMs, I focus on LLM applications research, agentic AI, and core LLM optimization problems
- In Reinforcement Learning, I am interested in LLM post-training using RL — I have published work on training generalist reasoning models using a novel GRPO-based approach at the AAAI 2026 PLAN-FM Bridge Workshop
- I worked as an Undergraduate Researcher at IIIT Delhi's FLaME.nlp Lab under Dr. Md. Shad Akhtar, with work published at the ACL 2026 Main Conference (Oral Presentation)
- I have had the opportunity to intern at the AI Institute, University of South Carolina (AIISC) under Dr. Amit Sheth, with work published at AAAI 2026 Main Conference, and at IIT Patna under Dr. Sriparna Saha
Peer-reviewed research at top-tier AI venues
| Paper | Venue | Type |
|---|---|---|
| 🧠 Measuring What Matters: Assessing Therapeutic Principles in Mental-Health Conversations | ACL 2026 Main Conference | 🎤 Oral Presentation |
| 📚 PAL: Personal Adaptive Learner | AAAI 2026 | 🪧 Poster Presentation |
| 💡 Rethinking Reward Models! A Conceptual Framework for Enhancing LLM Reasoning through Intrinsic Traits | AAAI 2026 PLAN-FM Bridge | 🎤 Oral Presentation |
🌍 Google Summer of Code 2026 — Open Source ML Contributor (Jun 2026 – Present)
- Developing semantic segmentation architectures (U-Net, SegNet) in PyTorch on multi-temporal satellite imagery, boosting coastline reconstruction accuracy by 16% across 1,000+ km of Alaskan territory
- Building longitudinal time-series forecasting models on large-scale remote sensing data arrays, reducing regional erosion predictive errors by 11%
🤖 Wadhwani AI — Machine Learning Scientist Intern (Jan 2026 – Jun 2026)
- Worked on improving domain-specific models for question answering using Retrieval-Augmented Generation (RAG)
- Developed a hierarchical chunking technique which improved retrieval and reranking results by 16% and 19% respectively on recall@k metrics
- Architected an enterprise agricultural knowledge base with Mistral OCR, orchestrating cloud-native ETL jobs via AWS S3 and GCP Cloud Functions to ingest 8,000+ multi-modal documents
- Deployed the end-to-end RAG pipeline (retrieval, reranker, and final answer generation) for commercial use using FastAPI + vLLM + Docker
🔬 AI Institute, University of South Carolina — Research Intern (Jan 2025 – Oct 2025)
- Engineered an agentic multi-modal AI pipeline achieving 94% factual accuracy in generating graded quiz assessments from lecture slides
- Built a personalized learning engine using hybrid reinforcement learning with Llama-3.2, boosting student retention by 19%
- Co-authored PAL: Personal Adaptive Learner, accepted at AAAI 2026
🧪 IIT Patna — Research Intern (May 2025 – Oct 2025)
- Pioneered a customized GRPO (Group Relative Policy Optimization) alignment architecture with a binary reward function for exact ground-truth matching, achieving a 27% exact-match accuracy boost on a complex cultural QA dataset
🏫 FLaME.nlp Lab, IIIT Delhi — Undergraduate Researcher (Jan 2024 – Jun 2025)
- Designed a novel knowledge distillation + Chain-of-Thought (CoT) architecture improving classification F1-score by 19%
- Curated a 15,000-record annotated dialogue dataset with automated quality-control, cutting label noise by 14%
- Co-authored paper accepted for Oral Presentation at ACL 2026 Main Conference; awarded Best Undergraduate Researcher
Generative AI & LLMs
MLOps & Cloud
Languages
Specializations
Deep Learning • Natural Language Processing • Reinforcement Learning
RAG Architectures • Agentic Systems • LoRA / QLoRA Fine-tuning
GRPO / PPO / RLHF • Chain-of-Thought Reasoning • vLLM • FAISS
Semantic Segmentation • Time-Series Forecasting