Bitun Banerjee
I build the platforms enterprises use to put AI agents into production: runtime, memory, tools and governance.
About
BioI lead product for the Agent Builder Platform in the CTO organization at J.P. Morgan Chase. My teams build the runtime, memory, tool integration and policy controls that developers across the firm use to build and run agents on shared identity, governance and observability.
I came to product as an engineer. I started out building energy systems at TechnipFMC in Houston, where I owned connected maintenance software and predictive failure models for offshore equipment. My MBA at NYU Stern and my years at Boston Consulting Group added the business lens: strategy, pricing and go-to-market.
I think in systems. An agent is only as good as the runtime, data, permissions and people around it. I bring that whole picture to AI-driven workflows, in plain language that engineers, executives and risk teams can all work from.
- Agent platforms & managed runtimes
- Agent memory, RAG & enterprise knowledge
- MCP & tool standards
- AI governance & entitlements
- Developer experience
- Capital markets & payments
The agent stack I run
J.P. Morgan · 2021 to nowAI Registry
find & reuseFirmwide directory of agents, tools and knowledge bases, with access tied to entitlements and responsible-AI policy. Playgrounds for building and evaluating agents.
Developer Experience
build & operateManaged Agents with durable execution, memory, human-in-the-loop checkpoints and guardrails. Agent Builder CLI and an AI studio for tracing, debugging and evals.
Platform
runAgent runtime and infrastructure, the Agent Builder SDK (skills, sandboxing, MCP-connected tools) and the Agent Context Engine, which joins memory, RAG, NL2SQL and knowledge bases into one context layer.
I reorganized the stack into these three product lines so the roadmap matches how developers actually consume it.
Experience
Most recent first-
2021 – now
J.P. Morgan Chase
Director, Product · Agent Builder Platform, CTO · New York
- Set the vision, roadmap and success metrics for the firm's agent stack, directing PMs and engineering partners.
- Open-sourced SmartSDK, JPM's in-house toolkit for building agents with RAG, document parsing and NL2SQL.
- Led roadmap and go-to-market for Quantitative Client Intelligence, an NLP markets data product that turns email and chat into trade signals using fine-tuned LLMs.
- Launched Cash Flow Intelligence, JPM's AI-driven cash management app for corporate clients.
-
2019 – 2021
Boston Consulting Group
Consultant · Texas and New Jersey
- Led product strategy for a commercial bank's digital payments business, including a data-driven sales model, cross-sell playbook and fintech partnerships ($200M upside).
- Due diligence for the merger of two of the world's largest turbine manufacturers.
-
2016 – 2017
Adcept Tech
Business Development Manager · Kolkata
- Brought IoT and IT solutions from US, Australian and Italian companies into India's public-sector energy and mining market.
-
2011 – 2016
TechnipFMC
Product Lead / Project Engineer · Houston
- Product owner for an always-online, API-enabled maintenance and service system that drove $100M in product sales.
- Built predictive analytics into offshore failure-mode detection, improving mean time to failure by 25%.
Education
NYU Stern School of Business
MBA, Analytics, Strategy & Finance, 2020. Named Faculty Scholar (full tuition), graduated with Distinction.
RV College of Engineering
B.E. Mechanical Engineering, 2011. First Class with Distinction.
Research & code
In progressPapers, experiments and open-source repositories on agent infrastructure. New work will be listed here as it is published.
SmartSDK
Toolkit for building agents from scratch, covering RAG, document parsing and NL2SQL. I led the decision to open-source it.
Notes on building agents for the enterprise
I write on Substack about running AI agents inside large organizations: context and memory, identity and governance, evaluation, and the product calls in between. Subscribe to get new posts by email.