I turn messy ERP, sales and operations data into trusted dashboards, automated reports and AI-assisted analytics that business teams actually use. At Walitechs Solutions I architect Microsoft Fabric + Power BI solutions across Sales, Finance, HR and Supply Chain, led a zero-data-loss Dynamics AX → D365 Business Central migration, and built an NL-to-SQL tool so non-technical managers can ask their data questions in plain English.
🏆 Employee of the Month, Walitechs Solutions
| 3+ yrs | 10+ | 50+ | ~50% | 35% | 0 |
|---|---|---|---|---|---|
| in data & MIS roles | enterprise Power BI dashboards | stakeholders served | less reporting time | faster SQL / Snowflake queries | records lost in the AX → D365 migration |
Every project runs on public or synthetic data, has tests and CI, and reports only numbers computed from its own data.
| Project | Business problem | Stack | Demo | Result (in the project) |
|---|---|---|---|---|
| NL-to-SQL Analytics Agent | Managers wait days for simple data answers | Claude API (tool use), DuckDB/SQL, sqlglot, Streamlit | ▶ Live demo (in-browser, no sign-up) | AST-based read-only guardrails, self-correcting agent, 32-question eval set, 66 tests (incl. DuckDB vs SQLite parity) |
| Fabric Lakehouse End-to-End | No trusted view of supplier OTIF and stock | Microsoft Fabric, OneLake, PySpark, Delta, Warehouse, Direct Lake | Deploy guide | Bronze → Silver → Gold with a DQ gate; OTIF 71.9% vs 85% target, 63 bad receipts quarantined |
| Snowflake JSON CDC Pipeline | Nested app events, duplicates, schema drift | Snowflake VARIANT, FLATTEN, Streams + Tasks, MERGE | make local |
Idempotent CDC; net sales match an independent recomputation to the paisa |
| ERP Migration Toolkit (AX → D365) + Fabric reporting | ERP cut-overs lose data quietly, and legacy AX reports break after go-live | Python, pandas, YAML specs, Microsoft Fabric (Lakehouse, Delta, Spark), T-SQL | Reconciliation report · Fabric AX views | 14/14 migration checks pass; BC → Fabric lakehouse with AX-compatible views, 10/10 reconciled |
| AI Report Automation | Hours lost writing monthly MIS commentary | Python, Claude API, Jinja2, GitHub Actions cron | Sample report | LLM narrative is number-checked against the KPI pack before it is e-mailed |
| Sales Intelligence (Power BI) | One sales truth + who is about to churn | SQL star schema, Power BI PBIP/TMDL, DAX, RLS, scikit-learn, SHAP | Power BI report | 2 report pages built from code, 33 DAX measures, calc group, dynamic RLS; churn model ROC-AUC 0.92 |
| Python Analytics Playbook | Slow, fragile notebook code in reporting jobs | pandas, Polars, DuckDB, Pandera, openpyxl, pytest | Benchmarks | 59% less memory, 415x faster than loops on 1M rows; idempotent ETL, Excel MIS, 31 tests |
| Power BI Business Dashboards | Dashboards that drive decisions | Power BI, theme JSON, DAX specs | 5 pages | Sales and Churn pages built in Power BI; Product, HR and Supply Chain designed next |
| DAX Measures Library | Re-writing the same DAX in every report | DAX | Docs | 30+ patterns with example outputs (YoY, Indian FY, Top N, RLS) |
Cloud & Fabric
BI
SQL
Python
ERP
AI
Engineering
- NL-to-SQL agent v2: Snowflake and Postgres connectors, few-shot retrieval from the eval set, a YAML semantic layer
- Fabric lakehouse on a live trial: pipeline run, Direct Lake report screenshots and Capacity Metrics
- dbt: moving the sales star schema to dbt models with tests and docs
Open to Senior Data Analyst, Power BI Developer / BI Architect, Analytics Engineer and GenAI Data Analyst roles in Gurugram / NCR, Bengaluru, Hyderabad or remote. 📫 shashan4321@gmail.com · 💼 LinkedIn · 🌐 Portfolio
🎓 MBA (Marketing & HR), GL Bajaj · B.Com (Hons), BHU · Data Analytics Certification, Ducat