AmritanshDixit

I sit with the people who'll use the software, build it end to end, and put AI agents to work in real delivery.

Software Development Engineer III at Citi, Model Risk Management
Offered a Devin Coach role, open to no more than 100 employees across Citi worldwide
Pune, India · About four years in enterprise and banking platforms
Java, Spring Boot, Angular, Python, MCP, Kafka
How I ship. Every module I own passes the same six gates. The ones moving through now are modules I've delivered.Released: 0

Stage 1 of 6

Requirements

I start with the people who will use it. For Assurance Review, that meant business users and the senior stakeholders who attest every model in the inventory twice a year.

SDE1 to SDE3Two consecutive promotions within two years
Citi Gold AwardLegal Entity module taken to production in 2–3 months
Devin Coach offerOpen to no more than 100 employees across Citi worldwide

Getting AI agents to do real work inside a bank

Citi rolled out Cognition's Devin coding agent to its developers. I use it in my own delivery, I've been offered a role coaching others on it, and I build with agents and MCP myself.

Citi · Devin coding agent

Devin in daily delivery, and an offer to coach others on it

In my SDE3 delivery, Devin takes the well-scoped tasks and a human reviews every change. That work led to an offer to become a Devin Coach, a role open to no more than 100 employees across Citi worldwide.

DeepLearning.AI with Anthropic

Model Context Protocol

Built an MCP server with tools, resources and prompt templates in FastMCP, tested it in MCP Inspector, wired a client into a chatbot and connected it to Claude Desktop.

View the course

Every day

LLMs in the loop

I use LLM assistants daily to investigate issues, draft designs and write prompts, and I keep the judgement calls with people.

Modules I've owned, and what happened after release

Most of my work lives inside Citi's Model Risk Management platform and its database infrastructure. Open any entry for what I did and the tools involved.

Legal Entity moduleModel Risk Management platform · Citi

In production within 2–3 months. Citi Gold Award.
  • Owned the module end to end: design, development, testing, deployment and cross-functional coordination.
  • Worked with product owners and QA to lock requirements early so the release date held.
  • Recognised with the Citi Gold Award for bringing it to production on that timeline.
JavaSpring BootAngularMicroservices

Assurance Review moduleModel Risk Management platform · Citi

The semi-annual model inventory review that senior stakeholders attest. Built solo, frontend to backend.
  • Twice a year, every model in the bank's inventory goes through an assurance and completeness review, and senior stakeholders attest to the result.
  • Owned the module alone: gathered requirements directly with business users, designed the solution, and built both the Angular frontend and the Spring Boot backend.
  • Took it from first requirement through to production release.
JavaSpring BootAngularREST APIsSQL

Calendar moduleModel Risk Management platform · Citi

A technically complex module, owned from design to deployment. Stakeholder appreciation.
  • Took ownership of one of the platform's more technically involved modules.
  • Delivered its design, development, testing and deployment.
  • Earned appreciation from stakeholders for the ownership shown through delivery.
JavaSpring BootAngular

MRM portal re-engineeringAudit Trail, AMR and OPA modules · Citi

Performance improvements and steadier platform behaviour across three modules.
  • Contributed to rebuilding the portal across the Audit Trail, AMR and OPA modules.
  • Worked with product owners, QA and senior engineers to resolve integration blockers.
  • Picked up the domain and codebase quickly and took ownership across workstreams.
JavaSpring BootAngularREST APIsSQL

Oracle database grid patchingCiti Technology Infrastructure · Oracle as a Service

A UI-driven patching feature adopted extensively across Citi, replacing manual steps.
  • Designed and built a UI for patching Oracle database grids.
  • Removed manual patching steps and the operational overhead that came with them.
  • The feature went on to wide use across the organisation.
JavaAngularOracleLinux

Infrastructure workflow modernisationCiti Technology Infrastructure

Better operational visibility and fewer manual interventions.
  • Helped move infrastructure workflows onto Autosys and automation tooling.
  • Made job status easier to see and cut the number of hand-run steps.
AutosysLinuxOpenShiftGit

PIR innovation projectCiti Graduate Program

Selected among the top 7 projects worldwide from all Graduate Program submissions.
  • Conceived the idea and carried it through to delivery in a competitive internal programme.
  • Combined product thinking with cross-functional collaboration to get it built.
Product thinkingPrototyping

Science reply botAxis intra-college hackathon · NIT Nagpur · 2021

A full-stack NLP bot that drafts factual replies to science tweets. Hackathon winner.
  • Reads a user's tweets, picks out the science-related ones and drafts a short, factual reply to each.
  • NLP pipeline in spaCy and NLTK: counts science terms, pulls context from Wikipedia and ranks candidate sentences by BLEU similarity to the tweet.
  • Flask API, React frontend with login and reply history in PostgreSQL, deployed on Heroku.
  • View the code on GitHub
PythonFlaskspaCyNLTKReactPostgreSQLHeroku

Dynamic pricing for smart gridsFinal-year project · NIT Nagpur · 2023

A reinforcement learning agent that sets electricity prices as demand shifts.
  • Built a reinforcement learning agent in Python that adjusts electricity prices to balance supply and demand on a smart grid.
  • Open the notebook
PythonReinforcement learningGoogle Colab

Three levels in two years at Citi

I joined Citi in 2023 and moved from SDE1 to SDE3 through two consecutive promotions, taking on more of the delivery each time.

SDE1C9 · Citi Technology Infrastructure, Oracle as a Service

Building for infrastructure teams

  • Designed and built UI-based Oracle grid patching.
  • Modernised workflows with Autosys and automation.
  • Shipped production-grade tools with cross-functional teams.
SDE2C10 · Model Risk Management · from Jul 2024

Re-engineering a regulatory portal

  • Rebuilt Audit Trail, AMR and OPA modules.
  • Unblocked integrations with product and QA.
  • Owned workstreams in a new domain.
SDE3C11 · Model Risk Management · from Jul 2025

Leading delivery end to end

  • Uses Devin in daily delivery; offered a Devin Coach role.
  • Solely owned the Assurance Review module, Angular frontend and Spring Boot backend.
  • Requirements, estimates, solution design and sprint plans through to release.
  • Coordinate 6–10 engineers and stakeholders; run sprint ceremonies.
  • Mentor 3–5 junior developers through reviews and knowledge sessions.
2023 – present · Citi, PunePeer recognition for running sprint ceremonies and aligning stakeholders

What I work with, and what I've been recognised for

Full-stack enterprise engineering and agentic AI, with the planning and people side of delivery that regulated platforms need.

Technical skills
AI and agentsDevin coding agent, LLM prompting, Model Context Protocol (FastMCP, MCP Inspector, Claude Desktop)
Languages and frameworksJava, Python, TypeScript, JavaScript · Spring Boot, Flask, Angular, React
Messaging and schedulingApache Kafka, Autosys
Infrastructure and DevOpsOpenShift, Git, Linux, Heroku
Machine learningReinforcement learning, spaCy, NLTK
Databases and APIsSQL, PostgreSQL, Oracle, REST APIs, microservices
DeliveryJira, Agile and Scrum, sprint planning and estimation
PracticeTechnical solution design, microservices architecture, regulatory module development, cross-functional collaboration

Recognition

  • Devin Coach offerOpen to no more than 100 employees across Citi worldwide
  • Citi Gold AwardFor taking the Legal Entity module to production in 2–3 months
  • Top 7 worldwideCiti Graduate Program innovation project, out of all global submissions
  • Two consecutive promotionsSDE1 to SDE3 within two years
  • Axis intra-college hackathon, winnerNIT Nagpur, second year

Education

B.Tech, Computer Science and Engineering

National Institute of Technology, Nagpur · Class of 2023

Final-year project: reinforcement learning for dynamic pricing in smart grids. Won the Axis intra-college hackathon in second year.

Need someone to make AI and software work inside a real organisation? Let's talk.

dixitamritansh@gmail.com
LinkedIn GitHub
LinkedInin/iamritansh
Based inPune, India
Open toForward deployed engineering roles