Bengaluru, India · MSc Data Science · AI / ML · Analytics Engineering
Open to data science and AI/ML opportunities
Helping teams turncomplex data intorigorous, scalable AI systems.
I build machine learning workflows, analytics pipelines and validation-first decision systems that can move from experimentation to real-world use without losing clarity or rigor.
Data Science · AI / ML · Bengaluru
Sanskar Swarup Das
Data scientist, ML builder and analytics-first problem solver.
Data scientist, ML builder and analytics-first problem solver.
I’m based in Bengaluru and focused on building dependable machine learning systems, smarter analytics workflows and AI-driven tools that stay understandable to the people using them.
What I enjoy most is the space between research and execution: benchmarking models, validating assumptions, improving pipelines and making sure the final output is genuinely useful rather than technically impressive only on paper.
Outside project work, I’ve also spent time teaching and mentoring through workshops on ML, generative AI and model evaluation — which has made communication one of the strengths I bring into technical teams.
REVA University, Bengaluru
Master of Science in Data Science
CGPA: 9.27 / 10Dharanidhar University, Kendujhar
Bachelor of Science in Computer Science
CGPA: 9.1 / 10REVA UniversityCognifyz TechnologiesTuSimpleDharanidhar UniversityAzure AI FundamentalsGoogle CloudLangChainPyTorchSQLPower BI
Built with
Machine learning, data infrastructure and validation discipline.
PythonRSQLPyTorchTensorFlowscikit-learnLangChainTransformersRAGGenerative AIPySparkBigQueryAzure AI-900GCPTableauPower BIEDAFeature EngineeringAnomaly DetectionETL Pipelines
Selected work
Research, systems and analytics workbuilt to stay reliable under pressure.
Selected research
An award-winning forecasting framework built to generalize across pathogen types without disease-specific training data.
Architected and validated a scalable deep learning time series forecasting framework using PyTorch and transfer learning, then stress-tested its assumptions and reliability under data shift.
View case study →
Selected project
Predictive analytics and reinforcement learning for more efficient food allocation across institutions.
Built predictive models using logistic regression, NLP and reinforcement learning, then paired them with ETL design and forecasting logic to improve allocation decisions.
- Logistic Regression
- NLP
- Reinforcement Learning
- Econometric Forecasting
Selected project
A real-time anomaly detection and alerting system built for safety-critical monitoring in rural households.
Built a real-time statistical anomaly detection system with Azure IoT integration, threshold-based risk logic and automated multi-channel alerts.
View case study →
What I build
Not just models. Systems that connect data, evaluation and decision-making.*
Machine learning systems
End-to-end model design, training, benchmarking, validation and iterative improvement using scikit-learn, PyTorch, TensorFlow and strong statistical reasoning.
Analytics engineering & ETL
Production-minded Python and SQL workflows with data-quality checks, anomaly detection and safer downstream reporting inputs.
LLMs, RAG & intelligent applications
Experience with LangChain, NLP, transformers, retrieval workflows, generative AI and agentic system patterns.
Technical thinking people can actually use
From workshops to technical writeups, I turn model behavior, trade-offs and limitations into clear language for teams and stakeholders.
* The best work is not only accurate. It is explainable, scalable and usable by the people making decisions.
How I evaluate work
Beyond scoreboards and headline metrics
My process focuses on whether a system can be trusted, maintained and acted on — not just whether it performs well once.
Benchmarking
Performance only matters in context, so I compare against baselines and alternatives before calling a model useful.
Robustness checks
Assumptions, failure modes and behavior under data shift matter just as much as accuracy on the primary dataset.
Data quality discipline
Upstream validation, anomaly checks and clean ETL design reduce the risk of bad decisions made on noisy inputs.
Actionable output
Results should help a team decide what to do next, not just report what a model achieved in isolation.
Experience
Projects, internships and leadership workthat show how the thinking lands.
85%
Built and validated ML systems at Cognifyz Technologies, then improved reliability through comparative testing, metric discipline and iterative refinement.
01
Dec 2025 — Jan 2026
02Cognifyz
Designed, trained and validated end-to-end ML models using scikit-learn and PyTorch, then benchmarked them against baselines with AUC-ROC, F1-score and RMSE.
Open case study →Jul 2025 — Sep 2025
03TuSimple
Engineered Python ETL pipelines with data-quality validation and anomaly checks, then optimized SQL queries to improve reporting speed and reliability.
Open case study →May 2025 — Dec 2025
AI & Analytics Club
Delivered technical workshops on ML, generative AI and statistics while mentoring students in Python, R and model evaluation.
Open case study →Tooling
Methods, frameworks and infrastructureI work with regularly.
- Linear & logistic regression
- Time series modeling
- Hypothesis testing
- Econometric analysis
- Model validation & benchmarking
- Numerical analysis
- scikit-learn
- PyTorch
- TensorFlow
- NLP & transformers
- Zero-shot learning
- Transfer learning
- Generative AI, RAG & LLMs
- Agentic AI with LangChain
- ETL pipelines
- PySpark
- Data validation
- Anomaly detection
- Feature engineering
- BigQuery
- MySQL & MongoDB
- Pandas & NumPy
- Matplotlib & Seaborn
- Power BI
- Tableau
- Azure AI-900
- Google Cloud Platform
- Azure IoT
- Bash, Git & Jupyter
What working with me feels like
Clear ownership, fast learning and thoughtful communication.
Fast ramp-up
I learn the domain quickly, identify the actual bottleneck and get productive without needing excessive hand-holding.
Ownership
I do not treat models like isolated notebooks — I think through data flow, reliability, documentation and handoff.
Communication
I can switch between technical depth and plain language, whether the audience is a recruiter, mentor, team or stakeholder.
Teaching mindset
Workshop and mentorship experience makes me strong at knowledge transfer, collaborative problem-solving and team learning.
Ready when you are
What do you need?
Whether it’s an ML prototype, data pipeline, dashboard, research implementation or AI product exploration, I’m open to conversations around roles, freelance work and collaboration.
Machine LearningAnalytics EngineeringETL PipelinesGenAI / RAGDashboardsResearch Projects