Gen AI Engineer
UMass Amherst | Aug 2025 – Present | Live App Demo
- Designed EduNotes, a multi-agent RAG study assistant centralizing study materials into a searchable semantic knowledge base that users query and build over time
- Content classification, adaptive summarization and notes generation based on input type with dynamic query routing
- Built multi-stage web search agent via DuckDuckGo and Google fallback with AI-driven query refinement at each step
- Orchestrated via LangChain & ChromaDB, to produce AI-generated flashcards, quizzes, notes, summaries, and progress analytics
- Research Mode with vision model analysis of PDF figures, tables, equations along with references from OpenAlex, arXiv & Semantic Scholar
- AI Chat agent with 7 modes (Socratic Tutor, Compare Topics, Research Writer, etc.) powered by Gemini 2.5 Flash and searchable conversation history
- Distributed workload across Llama 3.3 70B for generation, Llama 3.1 8B for lightweight tasks, and Llama 4 Scout for vision analysis
- Built with FastAPI & Streamlit, supporting 6+ input formats (PDF, URL, text, topic, web search) and 3+ output formats
Graduate Student Researcher
Cisco | Jan 2025 – Jul 2025 | Amherst, MA
- Built PerFine, an Agentic RAG framework for LLM personalization using LangChain
- Graph-based retrieval with Pinecone, FAISS, MCP for profile-grounded feedback
- 13% improvement in personalization, 10% in Meteor score over baselines
- Evaluated using LLM-as-a-Judge (G-Eval) on Yelp, Goodreads, Amazon datasets
Research Assistant
UMass IESL Lab | Aug 2024 – Dec 2024 | Amherst, MA
- Research under Prof. Andrew McCallum
- Built autoregressive model with lookahead decoding in superposition
- Efficient token search using just two forward passes with cross attention
- 15% improvement in BLEU score for MT5 machine translation
Machine Learning Engineer
Carelon Global Solutions (Elevance Health) | Jun 2021 – Jul 2023 | Hyderabad, India
Conversational AI System:
- Conversational AI bot with DistilBERT for intent classification & BERT QA for extractive question-answering
- Integrated RAG from S3 on SageMaker, Kubeflow; resolved 65% of patient queries across 1000+ daily interactions
Smart Recommendation Engine:
- Built healthcare recommendation systems using NER, SpaCy, XGBoost, LightGBM in PySpark
- 75% improvement in NDCG@5 for care plan recommendations
- REST APIs with Flask, Hive, MongoDB, Redis for end-to-end model automation
- Reduced care plan creation by ~2 hours through automated recommendations
- Created Splunk Dashboard for tracking KPIs from the user feedback
- Performed A/B testing with feedback-based model tuning drove 60% performance improvement
Sentiment Analysis Pipeline:
- Implemented Aspect-Based Sentiment Analysis on call transcripts using RoBERTa, SpaCy
- Provided actionable insights on patient satisfaction with 85% accuracy, 0.81 F1-score
Healthcare Knowledge Platform:
- Web-scraped healthcare articles indexed in Elasticsearch with ranking optimization; ~120ms retrieval
- Reduced information search time by 90%, enabling faster access to medical guidelines and policies
Production ML Infrastructure:
- Deployed to ENSO ML pipeline using RabbitMQ, Kubernetes, Kafka
- 70% reduction in deployment time via CI/CD pipelines
- Lambda functions for IBM to S3 to DynamoDB pipeline via AWS Glue
- ETL with Google Cloud Vertex AI, Airflow, Docker
AI Engineer Intern
SensorDrops Networks (STEP at IIT Kharagpur) | Aug 2020 – Sep 2020
- Real-time social distance monitoring during COVID using YOLOv3
- 90% detection accuracy with live video feed and bounding boxes
- Deployed on AWS EC2 with Docker, ~200ms latency
- Web interface for 4-camera live streaming with daily/weekly/monthly metrics
AI Engineer Intern
C-DAC, Pune | May 2020 – Aug 2020
- Deep CNN for COVID chest X-ray classification
- 92% accuracy on 3-class classification (COVID, non-COVID, normal)
- F1-score of 0.9 on HPC infrastructure
- CNN-based semi-supervised learning with VOS/VOT improvements