Skip to content
View sumankrsh's full-sized avatar

Block or report sumankrsh

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
sumankrsh/README.md
Suman Mukherjee, Data & AI Delivery Leader. I take AI from POC to P&L. 12+ years taking GenAI, ML and cloud-data programmes into governed production across insurance, life sciences and retail.

Ask my AI anything at mukherjeesuman.com  LinkedIn, 5K+ followers  Email me, replies in 24 hours  Read the blog

12+ years in AI and data. $5M Data and AI portfolio owned. About 30% lead conversion lift in pilot markets. 500TB+ cloud data migrated, about $280K saved per year.

I run Data & AI portfolios in insurance.
The part I care about most is what happens after the pilot.

Getting GenAI and agentic AI into production with the controls a regulator would expect:
clear ownership, audit-ready documentation, a human in the loop, and outcomes measured in conversion, retention and run-cost.

“I don't do demos. I ship outcomes, with governance that holds up in the audit room.”


✦ How I take AI from POC to P&L

Five steps: the demo (most AI dies as a POC), the data (profiling, data-quality checks, entity resolution), guardrails (Responsible AI, model risk, human in the loop), production (governed, audited, owned) and P&L (measured in conversion, retention and run-cost).

✦ Selected programmes

Enterprise AI transformation in Gulf insurance with a ~30% lead conversion lift. An AI/ML Centre of Practice for LLMOps and Responsible AI. Pricing analytics across 14,000+ retail stores. Compliance ML across 12 APAC markets lifting precision from ~70% to ~95%. A 500TB+ AWS migration saving ~$280K a year. AML detection and Power BI reporting for 200+ users.

Client and employer names are left out on purpose. Want the detail behind any of these? Ask my site: it answers from my real work.

✦ Built in the open

Ask-me-anything portfolio at mukherjeesuman.com Medicine and Drug Interaction Advisor

Global AI-Accelerator Toolkit Sentiment analysis with BERT

More: Recommendation system · Anomaly detection · Tabular deep learning

✦ Fresh builds

  • model-cost-pilot · Send each kind of request to the cheapest model that still passes your checks.
  • agent-crashlab · Break your agent's tools on purpose, and count what it does twice.
  • ai-policy-lint · Your AI policy as rules, checked on every agent config in every pull request.
  • prompt-ci · Prompts with owners, versions and tests, checked in every pull request.
  • agentbreak · Find out whether an attack makes your agent act, not just whether it says something odd.
  • Enterprise Data Skills · Five everyday data checks your team and your agents can run the same way, offline.

✦ Toolbox

Python, PyTorch, TensorFlow, scikit-learn, Azure, AWS, GCP, Docker, FastAPI, Git, GitHub Actions, Vercel

GenAI and agents: Azure OpenAI, AWS Bedrock, LangChain, RAG, vector search, agentic AI, n8n. ML and analytics: Python, SQL, PySpark, MLflow, Azure ML, propensity models, Power BI, Tableau. Cloud and data: Azure, Databricks, Synapse, AWS Redshift, AWS Glue, Snowflake, GCP. Governance: Responsible AI, LLMOps, model risk, human in the loop, audit-ready documentation. In the lab now: LangGraph, CrewAI, Claude Agent SDK, OpenAI Agents SDK, MCP, A2A, LLM evals, Ollama, Qdrant, Azure AI Foundry, ISO/IEC 42001, NIST AI RMF, EU AI Act.

✦ Credentials

Microsoft Certified Azure Solutions Architect Expert. TOGAF 9.2 Enterprise Architect. Certified AI Professional (CAIP). PMP in progress. PGP in AI and Machine Learning (UT Austin McCombs with Great Learning). PGP in Data Engineering (UMass with Simplilearn). MBA in Finance and Systems (SRM University).

✦ Latest writing

All posts →

✦ Off the clock

🎸 Guitar after hours  ·  🤘 Dream Theater, always  ·  🐾 Proud cat dad to Golu


Have an AI programme in mind? Let's talk. AI transformation, GenAI and agentic delivery, AI governance and advisory. I reply within 24 hours.

contactme@mukherjeesuman.com  ·  LinkedIn  ·  mukherjeesuman.com

Profile views
Views are my own and do not represent any employer.

Pinned Loading

  1. Sentiment-Analysis.ipynb Sentiment-Analysis.ipynb Public

    Fine-tuned BERT vs a TF-IDF + Naive Bayes baseline for classifying airline complaint tweets (Colab notebook).

    Jupyter Notebook 8

  2. ai-accelerator-toolkit ai-accelerator-toolkit Public

    Streamlit demo for planning enterprise AI programmes: readiness dashboard, cloud blueprints, ROI estimator and a GenAI playground, on synthetic data.

    Python

  3. Anomaly-Detection-Model- Anomaly-Detection-Model- Public

    Unsupervised anomaly detection on travel booking and ticket-cost records with PyCaret's Isolation Forest.

    Jupyter Notebook

  4. medicine-drug-advisor medicine-drug-advisor Public

    FastAPI drug-information and interaction checker for 34 common Indian and mental-health medicines, with an experimental RAG version.

    Python

  5. movie-recommendation-system movie-recommendation-system Public

    Movie recommender (TMDB and MovieLens data): popularity charts, content-based TF-IDF and SVD collaborative filtering.

    Jupyter Notebook