I build AI systems that can tell you when they don't know.
Things I have built
Four systemsEach one is a full system, not a demo — built end to end, with its own evaluation evidence. Open one to read how it actually works.
Multi-Agent LLM Evaluation Platform
Grades a chatbot's answers with a panel of three judges, then measures how much the judges actually agreed.
Secured Enterprise RAG Gateway
Answers questions over a private security corpus with citations — and refuses when the sources don't support an answer.
Clinical Facial Symmetry Measurement
Turns a dentist's by-eye judgement of the occlusal plane into an on-device angle measurement — offline, encrypted, in seconds.
Pediatric Dental Anxiety Assessment
Measures how anxious a child is before dental treatment — from four signals captured on a phone, in minutes.
Peer-reviewed
First authorEvaluating Chatbots in the LLM Era: A Hybrid Survey and Qualitative Synthesis of Evaluation Frameworks
Journal of Ambient Intelligence and Humanized Computing (Springer Nature) · Scopus Q1, Volume 17, 2026
DOI 10.1007/s12652-026-05091-6
The survey behind SIRIUS. Writing it is what made the gap obvious: everyone measures quality, almost nobody measures whether the measurement agreed with itself.
What I work with
- Programming languages
- PythonSQL
- LLM & AI systems
- Large Language Models (LLMs)Agentic AIMulti-Agent SystemsLangGraphLangChainRetrieval-Augmented Generation (RAG)Hybrid Retrieval (Vector + BM25)QdrantCohere RerankLangSmithPrompt Engineering (CoT, ReAct)Structured Outputs (Pydantic, JSON Schema)
- Machine learning
- TensorFlowKerasNumPyPandasData ProcessingModel Evaluation
- Backend
- FastAPIREST APIsSQLAlchemyPostgreSQLAlembic
- Frontend & mobile
- FlutterRiverpod
- Security
- JWT AuthenticationLlama Guard
- Infrastructure & testing
- DockerGitHub Actionspytest
- AI developer tools
- Claude CodeCodexAI-assisted Software DevelopmentAI-Powered DevelopmentRapid Prototyping
- Tools
- GitVS CodeJupyter NotebookGoogle ColabPyCharm
Most of these were learned by shipping one of the four systems above, not from a course.
Education
Wadia College of Engineering, Pune
B.E. Computer Engineering
Nov 2022 — Jun 2026
8.91
CGPA
Certifications
Machine Learning Specialization
DeepLearning.AI & Stanford University
Deep Learning with Keras and TensorFlow
IBM