Hi, I'm
CS (AI & ML) undergrad at Manipal University Jaipur, specialising in data analytics and applied machine learning. Experienced in demand forecasting, causal inference and process mining, with published research and three internships.
Download ResumeDownload Resume
SQL
Power BI
Advanced Excel
Pandas
NumPy
Plotly
Matplotlib
SciPy
SAP Analytics
PM4Py
ETL Pipelines
JupyterSQLPower BIAdvanced ExcelPandasNumPyPlotlyMatplotlibSciPySAP AnalyticsPM4PyETL PipelinesJupyterSQLPower BIAdvanced ExcelPandasNumPyPlotlyMatplotlibSciPySAP AnalyticsPM4PyETL PipelinesJupyterPython
TensorFlow
PyTorch
Scikit-learn
LangChainLangGraph
Hugging Face
OpenCVNumPyPandas
MLflowPythonTensorFlowPyTorchScikit-learnLangChainLangGraphHugging FaceOpenCVNumPyPandasMLflowPythonTensorFlowPyTorchScikit-learnLangChainLangGraphHugging FaceOpenCVNumPyPandasMLflow
React
TypeScript
JavaScript
Next.js
FastAPI
Flask
Spring Boot
Streamlit
Docker
GitHub Actions
Java
REST APIsReactTypeScriptJavaScriptNext.jsFastAPIFlaskSpring BootStreamlitDockerGitHub ActionsJavaREST APIsReactTypeScriptJavaScriptNext.jsFastAPIFlaskSpring BootStreamlitDockerGitHub ActionsJavaREST APIs
PostgreSQL
MySQL
MongoDB
Redis
Qdrant
SQL
Vercel
Render
Postman
Git
LinuxJupyterPostgreSQLMySQLMongoDBRedisQdrantSQLVercelRenderPostmanGitLinuxJupyterPostgreSQLMySQLMongoDBRedisQdrantSQLVercelRenderPostmanGitLinuxJupyterTech Stack
About Me
CS (AI & ML) undergrad at Manipal University Jaipur, specialising in data analytics and applied machine learning. Experienced in demand forecasting, causal inference and process mining, with published research and three internships.
Projects
Real-time alert correlation engine using TF-IDF vectorisation and time-windowed DBSCAN clustering to collapse hundreds of raw infrastructure alerts into root-caused incidents. Validated at 91.7% incident detection and 91.4% cluster purity across 8 independent runs on LogHub HDFS_v1 and AIOps Challenge 2020. Led Team SpaceX to 2nd place at the HPE SYNERGY 2026 hackathon.
Supply-chain analytics platform: LightGBM demand forecasting hitting 87.5% accuracy on the M5 (Walmart) Kaggle benchmark, with SHAP-explained drivers behind every number. 7 LangGraph agents cover forecasting, inventory, supplier risk and reporting; 9 production modules with Power BI dashboards and a human approval gate before any recommendation acts.
First unified framework bridging OCEL 2.0 Object-Centric Process Mining with Structural Causal Models. PC Algorithm with Fisher’s Z test recovered the causal DAG at F1 = 1.000 on planted ground truth, correcting naive correlation’s ~19% overestimation of supplier delay effects. Mixed SCMs per node (Logistic/GBM/Linear), validated across 10 seeds × 6 confounder strengths, on BPI Challenge 2019 data.
Production RAG system with LangGraph tri-routing across document index, web search and direct LLM. Measured against naive RAG on RAGAS: Faithfulness +7% (0.759→0.816), Context Precision +14% (0.875→1.000). Multi-provider fallback pool (Groq→Gemini→Cerebras→Cohere) recovers mid-stream, and a relevance fast-path skips query rewriting to save 1–2s per call.
Causal congestion platform built for Bengaluru Traffic Police, Gridlock Hackathon 2.0 (Flipkart × BTP). Proved parking violations causally lead congestion via DoWhy SCM backdoor adjustment. Novel PCIS metric, LightGBM + DBSCAN clustering, SHAP explainability, and a MILP officer-allocation optimizer across 54 BTP stations. End-to-end Streamlit dashboard.
Airbnb-style hotel booking platform, Java Spring Boot 3 + Next.js 15. Race-condition-safe inventory via JPA pessimistic write locks (SELECT FOR UPDATE) and two-counter model. JWT rotation (15-min/180-day), RBAC, Stripe webhook-confirmed booking, Docker multi-stage deploy, 365-row batch inventory auto-creation per room. Live on Vercel + Render.
Production-ready online food delivery application featuring a Customer Storefront, Admin Dashboard, and a scalable Spring Boot Backend. Secure JWT auth, AWS S3 image uploads, Razorpay payment integration, and a decoupled React UI.
Research
Peer-reviewed · Independent evaluation · Real datasets
Conversational Expense Tracker: A Multimodal AI System for Automated Financial Management
A hybrid twin-LLM routing system on WhatsApp that dispatches each query between a lightweight and a heavyweight model based on complexity, combined with NLP, OCR, and voice-to-text for automated expense logging and conversational analytics. The routing layer delivered 87% faster responses at 48% lower compute cost. Peer-reviewed and published at SSIC 2025, IIIT Allahabad.
Benchmarking Vision Transformers for Multi-Class Skin Lesion Classification on HAM10000
Comprehensive benchmark of 5 deep learning architectures (Swin Transformer, ConvNeXt Tiny, MobileNetV2, NFNet, and CNN+Attention hybrid) on the HAM10000 dermoscopic dataset (10,000+ images, 7 classes). Swin Transformer achieved 97.15% accuracy (±0.21%) via 5-fold stratified cross-validation. Research conducted during internship at IIIT Allahabad.
Credentials
Certifications · Achievements · Leadership
- Data Analytics EssentialsCisco Networking Academy · 2026
- Introduction to Data ScienceCisco Networking Academy · 2026
- SAP Analytics CloudSAP Learning
- SAP Signavio — Process MiningSAP Learning
- AI-900: Azure AI FundamentalsMicrosoft
- Deep LearningNPTEL Elite · IIT Ropar
- Machine LearningNPTEL Elite · IIT Madras
- Introduction to Modern AICisco Networking Academy · 2026
- S/4HANA Production Planning (PP)SAP Learning
- Business Processes in S/4HANA PPSAP Learning · Digital Badge
- Sourcing in S/4HANA (MM)SAP Learning · Digital Badge
- DSA CertificationCodeChef · 150+ solved on LeetCode
Journey So Far
Research Intern, Deep Learning & CV
IIIT Allahabad · Hybrid
Engineered TensorFlow/Keras pipelines for medical image classification on HAM10000 (10,000+ dermoscopic images, 7 classes). Benchmarked 5 architectures; Swin Transformer achieved 97.15% accuracy (±0.21%). Research manuscript under preparation for peer-reviewed publication.
General Secretary
Student Research Forum · Manipal University Jaipur
Leading MUJ's Student Research Forum, a campus-wide body promoting academic research culture. Organized research symposia, paper-writing workshops, and inter-department collaboration events. Coordinated events with 800+ cumulative attendees and mentored juniors on research methodology and publication.
Software Engineering Intern
Grids Apps LLC · Remote
Built and deployed a 19-route dental clinic platform (Booking, Patient Dashboard, Smile Assessment, Before & After) using a React + TanStack Router/Query frontend wired to a Laravel REST API backend, with a Filament admin panel seeding 162 appointments, 30 patients, and 8 services.
Research Intern, Machine Learning
IAESTE · VŠB-TU Ostrava, Czech Republic 🇨🇿 · Remote
Selected through the International Association for the Exchange of Students for Technical Experience (IAESTE) for a research internship at VŠB-Technical University of Ostrava. Beginning ML research within a university lab environment, one of ~120 IAESTE exchange students globally selected for Czech Republic placements.