Muhammad Junaid Iqbal. Senior Software Engineer

03.Problems I have solved

• Single sign on for a partner bank's membersMembers open the investing app from their bank dashboard, already signed in.Surmount and BlossomBackendFull stack

  Problem

  Blossom, a banking platform, wanted its members to reach investing from its own dashboard without a second login. Blossom had not written any code yet, and the design had to work for future partners too.

  What I did

  Built it on OpenID Connect with authorization code and PKCE: the partner signs the member in, Surmount trusts it. Each partner is set up by config, not code. Wired the landing in the Next.js app. In review I found the partner off switch was skipped for returning members, and fixed the order so turning a partner off really blocks access.

  • Django
  • OpenID Connect
  • PKCE
  • Next.js
  • TypeScript

• One backend for four investing brandsSurmount, Blossom, Quiver and Daly run on one Django backend with live trades.Surmount white labelsBackendFull stack

  Problem

  Each brand needed its own app, users and rules, but one team could not run four backends that place real trades.

  What I did

  Shipped features across the shared multi tenant Django backend, separated by organisation. Built the KYC and deposit flows with Plaid in the Blossom Next.js app, owned the Django migration graph and staging deploys on AWS ECS, and reviewed other developers' pull requests.

  • Django REST Framework
  • Celery
  • PostgreSQL
  • Next.js
  • Plaid
  • AWS ECS

• Keeping live trades in sync with the brokerSells that went through were no longer reported as failed.SurmountBackend

  Problem

  The code decided a sell failed when the broker client returned nothing. The client also returns nothing when a reply is lost, so real sells with real money raised failed alerts and support tickets.

  What I did

  Made the broker's order list the source of truth. When a reply is lost, the code checks what the broker actually did, and allows for the delay before a new order shows up. I also handled user tickets on stuck trades, deposits and withdrawals and fixed the causes.

  • Python
  • Django
  • SnapTrade
  • Alpaca

• Four LLM features shipped to real usersA medical condition checker, script generation, content moderation and a podcast generator.CodainerAIBackend

  Problem

  Each product needed an LLM feature that worked every time for app users, not a demo that works once.

  What I did

  Built agent workflows with LangGraph and the OpenAI Agents SDK on FastAPI and Django, with tool calls and replies streamed over SSE and WebSockets to the React Native apps.

  • FastAPI
  • LangGraph
  • OpenAI Agents SDK
  • SSE
  • WebSockets

• A talking 3D avatar with real time lip syncThe avatar's mouth moves in sync with the voice as it speaks.CodainerAIFrontend

  Problem

  A conversational avatar feels broken if the voice and the mouth do not match, and replies had to stream without long waits.

  What I did

  Used Azure Speech for text to speech and speech to text, and drove the lip sync from its viseme events in real time. Replies stream over WebSockets.

  • Azure Speech
  • Visemes
  • WebSockets
  • Next.js
  • Django REST

• Lead intake automation for a US law firmA new enquiry becomes a case, a booked consultation and an attorney alert, without manual steps.Immigration law firm, client workBackend

  Problem

  Every enquiry from the website form or phone was copied by hand into the case system, then a coordinator booked the meeting, told the attorney and sent follow ups.

  What I did

  Built it on Power Automate: enquiries become Docketwise cases through a custom connector, consultations get booked, the right attorney is notified in Teams, and Mailchimp follow up runs on its own. Checked every integration and licence before building.

  • Power Automate
  • Docketwise API
  • Microsoft Teams
  • Mailchimp

05.Problems I can take on

Work I want to do next. Each one builds on something I have already shipped.

• AI agents that call business APIs

  Proof: I build agents with LangGraph and the OpenAI Agents SDK, and on Surmount I worked with SnapTrade, Alpaca and HubSpot.

  AIBackend

• Tests and evals for LLM features

  Proof: I shipped four LLM features at Codainer, and I reproduce a bug before I fix it.

  AI

• Payment and ledger reconciliation

  Proof: I fixed sell reconciliation against the broker, and moved balance checks to Decimal.

  Backend

• RAG over company documents, with sources shown

  Proof: I built a chat tool over legal case files and streamed LLM replies over SSE and WebSockets.

  AIFull stack

• Cutting LLM cost and response time

  Proof: My four LLM features ran in production, where every token and every second of wait was paid for.

  AIBackend

• Multi tenant SaaS: auth, billing and data isolation

  Proof: One Django backend served four brands, I built partner SSO, and caught a billing bypass on Sapius.

  BackendFull stack

• Workflow automation across the tools a team already uses

  Proof: I connected a law firm's intake form, case system, Teams and Mailchimp on Power Automate.

  Backend

• Design systems and fast React frontends

  Proof: I built shared component libraries at Devorium and moved a codebase to Tailwind and a Turborepo monorepo.

  Frontend

• Taking a web product to iOS and Android

  Proof: I shipped four React Native apps to both stores, with in app purchases.

  FrontendFull stack