I build AI systems that automate complex enterprise workflows, from multi-agent pipelines to document processing. Research-backed, production-tested, and always shipping with measurable results.
LLM & GenAI
ML & NLP
Cloud & Infra
Engineering
Multi-agent pipeline (Claude, Gemini, Tavily) with a 3-phase writing chain and self-improving hooks. Aggregates 14 news sources into publish-ready content.
GPT-4.1 Vision PDF extraction supporting 12 document types with JSON schema enforcement. Handles invoices, medical records, contracts, and more.
Multi-LLM system (Claude + GPT-4o-mini + Gemini) for legal practice area classification. Consensus-based routing across 50+ practice areas.
Full-stack RAG with a 6-node LangGraph pipeline, dual-layer confidence scoring, and Mistral fallback. Built for legal document analysis at scale.
| Role | Company | Impact | Stack |
|---|---|---|---|
| AI Engineer | CrossingLegal | Reduced H-1B processing from 5 days to 10 minutes | AWS Step Functions, Lambda, Bedrock, Claude API, RAG, Titan Embeddings, Pydantic |
| AI Engineer Intern | Boredm LLC | Fine-tuned GPT for 58-entity NER, 99.1% accuracy | OpenAI API, Pydantic, Python, TypeScript |
| Research Assistant | WPI | Published at AMCIS 2025, 0.94 cosine similarity | GPT-4, Gensim, NLTK, scikit-learn, Playwright |
| AI Engineer Intern | SmartStream | 99% NER accuracy, 80% faster extraction | AWS Textract, SageMaker, Comprehend |
- AMCIS 2025 - Zero-shot summarization framework using GPT-4 for qualitative research synthesis. Achieved 0.94 cosine similarity with human-authored summaries.
- ICFAiSE 2025 - Water turbidity prediction using LSTM with attention mechanisms. R² of 0.98 on real-world environmental data.
Other Projects
- water-quality-forecasting-ml - Ensemble methods for lake turbidity forecasting
- environmental-data-projects - Environmental data analysis collection
- lake-turbidity-forecasting-ensemble - XGBoost + CatBoost ensemble
- bangalore-water-quality-ml - Urban water quality prediction
- lake-turbidity-analysis - Exploratory turbidity analysis
- time-series-regression-xgboost - Time series regression with XGBoost
- facial-expression-recognition-cnn-bilstm - Hybrid CNN-BiLSTM for emotion detection
- facial-emotion-recognition-v3 - Facial emotion recognition pipeline
- emotion-detection-deep-learning - Deep learning emotion classifier
- emotion-classification-lstm - LSTM-based emotion classification
- heart-attack-prediction-ml - Heart attack risk prediction
- human-activity-recognition-lstm - LSTM for sensor-based activity recognition
- sensor-data-classification-dl - Deep learning sensor classification
- nlp-sentiment-analysis - Sentiment analysis pipeline
- ml-classification-projects - Classification project collection
- web-scraping-projects - Web scraping utilities
- statistical-ml-analysis - Statistical ML analysis
- regression-diagnostics-ml - Regression diagnostics
- data-visualization-analysis - Data visualization projects
- Leetcode - DSA practice and solutions
Let's build something that ships.
