APPLIED AI ARCHITECT · MICROSOFT

Ritwick
Dutta.

I help enterprises turn AI ambition into capability.

I connect customer context, technical depth, and the people who make AI adoption real. My work spans GenAI, agentic systems, and the platforms, controls, and operating models behind them.

Portrait of Ritwick Dutta

Toronto, CanadaWorking globally

15+

years across
AI, cloud & data

21

published articles
on AI & leadership

07

selected projects
& workshops

EXPERIENCE ACROSS

Microsoft TELUS Bell Rogers Bellwether Investment Management

01Enterprise GenAI

The model harness, controls, and operating foundation are the real differentiator.

02Agentic systems

Move from brittle task automation toward governed, context-aware autonomy.

03Responsible adoption

Trust comes from testable evidence, evaluation gates, and human oversight.

04Customer solutions

Connect executive ambition, architecture decisions, and the teams who run them.

01 / ARTICLES & PERSPECTIVE

Thinking in public.
Learning in the open.

Source-cited technical perspectives on this site, alongside published articles on enterprise AI, agentic systems, and the human side of leadership.

ON THIS SITE / TRUSTWORTHY AI IN PRACTICE

Short answers, deeper reasoning, and explicit limits. Grounded in public work and dated reviews of Microsoft Learn guidance.

02 / SELECTED WORK

Proof, not just
a point of view.

Workshops, demos, and applied experiments, plus an active founder-built venture with a live product. A closer look at how I connect AI capability with the realities of building, operating, and adopting it.

Showing all 7 projects

03 / LEADERSHIP IN PRACTICE

Lead the direction.
Understand the details.

I work at the intersection of customer needs, technical decisions, and organizational capability. The aim is not simply to demonstrate AI. It is to help teams move forward with confidence.

01

Start with the customer

Translate business ambiguity into architecture choices. Bring customer teams, technical stakeholders, and product engineering into the same conversation.

Architecture advisory & design reviews at Microsoft
02

Build for the real world

Treat evaluation, safety, governance, and supportability as design requirements. Help teams understand what needs to be true before an AI workload moves forward.

Production-readiness reviews & operational workshops
03

Grow the capability

Combine hands-on delivery with team and portfolio leadership. Make AI understandable through practical enablement, curriculum development, and shared implementation patterns.

AI leadership at TELUS & Bell; curriculum development and technical instruction

A CAREER BUILT ON DELIVERY

Microsoft

Oct 2025 - Present

Applied AI Architect, Financial ServicesCURRENT

Advise regulated organizations on GenAI and agentic architecture, production readiness, and platform governance. Connect customer engineering teams with product groups through workshops and design reviews.

TELUS

Oct 2023 - Oct 2025

Lead AI Engineer

Led applied AI delivery across multimodal assistants, agentic problem management, and workforce forecasting. Built two RAG-based assistants for field technicians and sales enablement.

Bell

Nov 2021 - Sep 2023

Senior Manager, Artificial Intelligence

Owned AI and data strategy transformation and an enterprise ML portfolio. Advanced Google Cloud data foundations, GenAI personalization, and responsible AI practices.

Rogers Communications

Nov 2019 - Oct 2021

Manager, Data Scientist

Led cloud analytics transformation and a machine learning portfolio across sales, marketing, and pricing, including recommendations, segmentation, and NLP.

Bellwether Investment Management

May 2019 - Feb 2020

Quantitative Analyst, Portfolio Management

Worked on the quantitative construction of the Alpha and Balanced Adaptive ETF funds. Built a Python and Bloomberg BQNT backtesting engine with automated research reporting. Developed risk-return analytics and researched RNN, GRU, and LSTM return forecasting.

QUANTITATIVE ROOTS

Test it. Then trust it.

Work on the Alpha and Balanced Adaptive ETF funds taught me to treat every result as a hypothesis until it survives data it has not seen. I bring that habit to AI evaluation today.

04 / CONCEPTUAL DEPTH

Depth is more than
a list of tools.

The most useful technical leaders can move from first principles to implementation, then connect both to the decision in front of the room. Curriculum development and instruction have been a rigorous test of that range.

CURRICULUM DEVELOPED AND TAUGHT

Natural language processing, end to end.

A complete learning progression from statistical language foundations to modern pretrained models, with implementation, evaluation, interpretability, and responsible use treated as one connected discipline.

14-week curriculum and 10-module online sequence
  1. 01

    Language foundations

    NLU and NLG, lexical processing, regular expressions, normalization, corpora, N-grams, probabilistic language models, text classification, and information retrieval.

  2. 02

    Structure and meaning

    Sequence tagging, HMMs, MEMMs, Viterbi decoding, POS tagging, NER, chunking, semantic roles, word sense disambiguation, vector semantics, embeddings, and speech systems.

  3. 03

    Neural sequence models

    Feedforward networks, PyTorch, backpropagation, RNNs, bidirectionality, LSTMs, sentiment analysis, tagging, and text generation.

  4. 04

    Attention and transformers

    Encoder-decoder models, attention, self-attention, multi-head attention, positional encoding, transformer implementation, and interpretability.

  5. 05

    Pretrained models and responsible use

    BERT, GPT, BART, transfer learning, fine-tuning, classification, QA, NER, summarization, discourse, coreference, probing, ethics, and trade-off evaluation.

ADDITIONAL COURSE DOMAINS TAUGHT

Six adjacent domains that strengthen how I connect AI systems, data, security, modelling, and stakeholder decisions.

01

Emerging AI systems

Generative model foundations and the operating questions around new AI capabilities.

  • VAEs and GANs
  • Image and video generation
  • GenAI APIs
  • Deep reinforcement learning
  • Bias, deepfakes, privacy, and IP
02

Machine learning

Model selection from classical methods through neural approaches, with evaluation tied to real problems.

  • Linear and logistic models
  • PCA and feature selection
  • Naive Bayes and probabilistic classifiers
  • Decision trees, KNN, and clustering
  • Neural networks and model evaluation
03

Data analytics

The practical path from raw data to analysis, visual evidence, predictive models, and recommendations.

  • Python, NumPy, and Pandas
  • Data preparation and wrangling
  • Statistical visualization
  • Regression and model selection
  • Clustering and recommender systems
04

Database security

Defense in depth across identity, code, data, monitoring, and continuity.

  • CIA and threat foundations
  • Authentication, authorization, and least privilege
  • SQL injection and secure code
  • Row-level security, masking, and encryption
  • Defender for SQL, auditing, KQL, ledger, backup, and recovery
05

Statistical and predictive modelling

Choosing, improving, and explaining models for business decisions.

  • Regression and logistic regression
  • Discriminant analysis and regularization
  • Decision trees and random forests
  • Clustering and conjoint measurement
  • Ensembles, accuracy, and interpretability
06

Data storytelling and leadership

Turning analytical evidence into a decision that stakeholders can understand and act on.

  • Decision theory and framing
  • Stakeholder resistance to change
  • Visualization choices
  • Rhetoric, white papers, and data stories
  • Calls to action and executive presence

The value is not an institutional affiliation. It is the discipline of explaining complex systems clearly, testing understanding through implementation, and making technical trade-offs legible to others.

05 / A LITTLE ABOUT ME

A builder's mindset.
A leader's perspective.

Current role
Applied AI Architect, Microsoft
Base
Toronto · Global advisory
Focus
Enterprise GenAI, Agents & Governance

My work connects the rigor of architecture with the realities of people, products, and adoption.

I've built AI assistants, led enterprise AI and data portfolios, and advised regulated organizations on the path toward production-ready GenAI. Today, I bring that experience to customer architecture at Microsoft.

My background also includes quantitative investment management, including ETF portfolio research, along with analytics, curriculum development, and technical instruction. That mix keeps me interested in both questions: can we build it well, and will it actually help?

I also build and operate products of my own. Axolytica is an active, founder-built AI wellness venture with a live early-access product. It keeps me close to the decisions behind product direction, end-to-end delivery, and ongoing operation. Read the Axolytica brief

AI systems

GenAI & agents / RAG / Evaluation & safety / MLOps

Platforms

Azure / Google Cloud / Kubernetes / Enterprise data

Finance

Portfolio construction / Backtesting / Risk analytics / Return forecasting

Leadership

Customer solutions / AI strategy / Technical enablement

Master in Quantitative Investment Management, University of New Brunswick
MBA in Strategy, Amity Business School

LET'S CONNECT

Let's make
AI useful.

AI strategy. Customer solutions. Technical leadership.
Meaningful conversations start with a real problem.

Request a discussion rittwick.dutta@gmail.com

Based in Toronto. Conversations across North America and beyond.

Ritwick's AI portfolio assistant

Hello, welcome to Ritwick's Intellectual Companionship. Let me know what interests you.

Messages are processed by an AI service to generate replies and are not stored on this website. Please avoid sharing sensitive details. Replies may be incomplete.