Open to full-time roles · Lahore or remote

I build backend systems for a legal platform used by 300+ law firms, mostly the parts where LLMs read and write massive documents.

Associate Software Engineer at Devsinc, working on LawPractice.ai, which plaintiff law firms in the United States use. Before this I built frontends with Next.js and React.

See my workGet in touch
LinkedInGitHub
Muhammad Junaid
Associate Software Engineer (AI/ML) · DevsincBS Software Engineering · FAST-NUCESLahore, Pakistan · Remote is fine
Scroll down
.NET 8 & C#LawPractice.ai · 300+ US law firms
Next.js 14ResQ CRM · fleet dispatch
Azure OpenAIdemand letter drafting
FastAPI & Claude APIISO 13485 audit engine
Gemini 3.1 & 3.5Juno AI SSE streaming
Agentic RAGJuno AI in-memory vector matrix
TypeScriptevery full-stack project
PythonAI agents, NLP & microservices
Groq Whispervoice gold valuation advisor
Socket.IO & WebSocketsreal-time chat presence & rooms
.NET 8 & C#LawPractice.ai · 300+ US law firms
Next.js 14ResQ CRM · fleet dispatch
Azure OpenAIdemand letter drafting
FastAPI & Claude APIISO 13485 audit engine
Gemini 3.1 & 3.5Juno AI SSE streaming
Agentic RAGJuno AI in-memory vector matrix
TypeScriptevery full-stack project
PythonAI agents, NLP & microservices
Groq Whispervoice gold valuation advisor
Socket.IO & WebSocketsreal-time chat presence & rooms
Cloud Firestorelive GPS vehicle telemetry
Supabase & PostgreSQLMawaddah matchmaking portal
Prisma ORM & MySQLrelational schemas & migrations
Upstash Redissliding-window rate limits
RabbitMQasynchronous document queues
Docker & Linuxservice containerization
CI/CD & GitHub Actionsautomated build & test pipeline
Express & Node.jsREST APIs & auth middleware
Java & Hibernatepharmacy POS transactions
TensorFlow & KerasBiLSTM news classification
Cloud Firestorelive GPS vehicle telemetry
Supabase & PostgreSQLMawaddah matchmaking portal
Prisma ORM & MySQLrelational schemas & migrations
Upstash Redissliding-window rate limits
RabbitMQasynchronous document queues
Docker & Linuxservice containerization
CI/CD & GitHub Actionsautomated build & test pipeline
Express & Node.jsREST APIs & auth middleware
Java & Hibernatepharmacy POS transactions
TensorFlow & KerasBiLSTM news classification
+

US Law Firms

Active firms on LawPractice.ai

case study
%

Prep Time Cut

Medical chronology & demand drafting

case study
ms

Voice Latency

Groq Whisper audio transcription

case study
-Dev

Frontend Lead

Next.js rebuild with live fleet GPS

case study

SYSTEM SCHEMATIC // AGENTIC ARCHITECTURE

Production agent architecture behind Junaid's AI assistant. Backed strictly by verified code, guarded by Upstash Redis rate limits, and streamed via Server-Sent Events.

EDGE GATEWAY // PROTECTION
20 req/min · 200/day

Dual Sliding-Window Ingestion

Every request is checked at the edge using Upstash Redis sliding windows. Two limits (20 requests per minute and 200 per day per visitor) keep the assistant fast and prevent spam.

Upstash RedisSliding WindowRate LimitingEdge Protection
GATEWAY_STATUS: [VERIFIED] Upstash Redis rate limit armed · 0ms overhead
AGENTIC_PIPELINE.DAG
SSE STREAM ACTIVE
CIRCUIT_BREAKER: READY
01 // EDGE GATEWAYUpstash Redis Rate Limit20 req/min · 200/day sliding window02 // KNOWLEDGE ENGINEProfile Data & Knowledge13 Case Studies + 61 Repos03 // MODEL CASCADEGemini 3.1 Flash-Lite12s fast failover → 3 / 3.5 FlashThought signature preservation04 // TOOL EXECUTION BUSMulti-Round Agentic ToolsSemantic search · Live git · EmailMax 4 rounds · 3 calls per round05 // DISPATCH PROTOCOLServer-Sent Events (SSE)Sub-100ms TTFB · Status traces

01 / About

I'm an Associate Software Engineer at Devsinc in Lahore. My primary project is LawPractice.ai, an AI platform used by more than 300 plaintiff law firms in the United States. Most of my day-to-day work is backend development with ASP.NET Core, LLMs, and RAG—taking messy case files and medical records and turning them into structured documents like demand letters, which helped cut preparation time by about 70%. I also worked on YCG, an automated compliance auditor built with FastAPI and Claude to check quality manuals against ISO standards.

Before Devsinc, I interned as a frontend engineer at Kryptomind, building web interfaces with Next.js and React. Over time, I grew into full-stack development—leading a small team of three on a real-time CRM called ResQ, and building projects on my own like an e-commerce store with Prisma, a real-time chat app with WebSockets, and Mawaddah, a matchmaking platform with Supabase.

I graduated with a BS in Software Engineering from FAST-NUCES. Alongside my studies, I spent a lot of time volunteering and organizing events. At SOFTEC'23, I was Deputy Head of Marketing, helping lead a team of about 40 students and closing sponsorship deals that raised over PKR 1,000,000. I also volunteered in operations at Future Fest'23 and completed the Aspire Leaders Program. Organizing events and working with sponsors taught me how to communicate clearly and take ownership beyond just writing code.

I like to understand what users actually need before jumping into code. I prefer clean, straightforward solutions over unnecessary complexity, and I want whatever I build to stay reliable and easy to maintain long after the demo.

See my experience
Lahore, Pakistan · Remote is fine
BS Software Engineering · FAST-NUCES
AI/ML · Full-Stack · Platform Engineering

02 / Experience

I started in frontend, moved to backend, and ended up working on AI features.

Dec 2025 — Present

Associate Software Engineer

Devsinc · Lahore, Pakistan

  • Primary project: LawPractice.ai. Built backend features using ASP.NET Core, LLMs, RAG, and MCP. My work on processing legal demands and case summaries helped cut document preparation time by 70% and made demand letters about 7x faster to turn around.
  • Helped build the APIs, AI agents, custom tools, and RAG pipelines that read, extract, process, and generate documents, and improved their prompts. Together these helped cut documentation errors by about 90%.
  • Kept the OCR, document reading, and document writing pipelines running across different document types, including large medical records of more than 1000 pages, added new ones when needed, and fixed issues clients reported in production. The platform is now used by 300+ law firms.
  • Also contributed to YCG, an automated compliance auditing tool built with FastAPI, Streamlit, and Claude to review Quality Manuals clause-by-clause against ISO standards.
  • Day to day: ASP.NET Core, Microsoft Azure, Azure Document Intelligence, AI Search, Azure SQL, Cosmos DB, RabbitMQ, and Python/FastAPI.
ASP.NET CoreMicrosoft AzureAzure OpenAIVector Databases/StoresLLMsRAGAgentic AIMCPRabbitMQFastAPIPython
View offer letter
Oct 2025 — Dec 2025

Software Engineer Intern

Devsinc · Lahore, Pakistan

  • Worked on backend development, bug fixes, and AI features for LawPractice.ai.
  • That work led to my current full-time role.
BackendAI/MLClient collaboration
View offer letter
Aug 2024 — Nov 2024

Software Engineer Intern

Kryptomind LLC · Lahore, Pakistan

  • Built interfaces with animations and 3D models using GSAP and React Three Fiber across several projects, and used Lenis for smooth scrolling.
  • Connected Next.js, TypeScript, and React frontends to REST APIs. Server-side rendering and code splitting took one project's Lighthouse score from 55 to 90.
  • Worked on an NFT marketplace and learned Web3 basics: blockchain, smart contracts, and wallet integration.
Next.jsTypeScriptReactGSAPLenisFirebaseWeb3
View experience letter

Education

BS Software Engineering

FAST-NUCES · 2021 — 2025

Learning

AI, Backend, Frontend, and DevOps

Mostly self-paced courses on Udemy, Coursera, Github, and Youtube where I built projects along the way

03 / Selected Work

A look at systems I've built, the problems I ran into, and how I solved them.

01
ProductionAI & LLMs

LawPractice.ai

DocReader and Demand Drafting pipelines for 300+ US plaintiff law firms.

Processes 1,000+ page medical dossiers with Azure Document Intelligence OCR to extract treatment chronologies and generate cited settlement demand letters.

300+Law Firms
70%Prep Time Cut
7xTurnaround
.NET 8ASP.NET CoreBlazor WebAssemblyAzure OpenAIAzure Document Intelligence
Case study
Live demo
02
ProductionWeb & Full-Stack

ResQ CRM

Frontend rebuild from scratch of an emergency disaster recovery and insurance response CRM with live fleet tracking.

Led the frontend rebuild from scratch of an emergency fire and disaster recovery CRM at Kryptomind, replacing an outdated legacy system with real-time GPS crew tracking via Cloud Firestore.

Live GPSFleet Map
FirestoreCost-Effective Sync
Next.jsDashboard
Next.jsTypeScriptNextAuth.jsCloud FirestoreTailwind CSS
Case study
03
ProductionAI & LLMs

Juno — AI Portfolio Assistant

Live portfolio assistant with Server-Sent Events, in-memory Agentic RAG, and GitHub tool-calling.

Answers questions about my work and inspects public repositories in real time using streaming SSE responses with in-memory Agentic RAG, automated multi-model fallback, and Upstash Redis rate limits.

Live on SiteProduction
Agentic RAG< 0.2ms Vectors
Dual LimitUpstash Redis
Next.js 14TypeScriptAgentic RAGGemini 3.1 & 3.5In-Memory Vector Search
Case study
Live demo
04
Production (Internal)AI & LLMs

YCG — ISO Compliance Document Auditor

Internal compliance auditing platform checking Quality Manuals against ISO 13485:2016.

Audits standard operating procedures clause-by-clause using multi-stage Claude prompt pipelines, generating gap analysis reports on a Streamlit dashboard.

ISO 13485Standard Audited
Multi-StagePrompt Pipeline
FastAPI + UIArchitecture
FastAPIStreamlitPythonClaude APIJSON Schema
Case study
05
Client ProjectWeb & Full-Stack

Mawaddah

Full-stack matchmaking portal with relational preference search and Supabase services.

A private matchmaking platform that replaced manual Google Forms with multi-preference candidate filtering, role-based access, and Supabase backend services.

FreelanceClient Project
SupabaseBackend Service
350ms → 120msQuery Optimization
Next.jsReactTypeScriptSupabasePostgreSQL
Case study
06
Personal ProjectAI & LLMs

Gold Investment Estimations Assistant

Voice-enabled precious metals calculator and market advisor on Hugging Face Spaces.

Converts voice queries into gold valuations across regional units and purities via Groq Whisper, paired with deterministic math and Gemini advisory.

HF SpacesCloud Hosted
Groq WhisperVoice Input
Gemini APIAdvisory Model
PythonGradioGoogle GeminiGroq WhisperOpenAI SDK
Case study
Live demo Source
07
Academic / CourseworkWeb & Full-Stack

Promptopia

Coursework training project exploring the Next.js App Router and NextAuth.

A prompt discovery app with Google OAuth authentication, tag-based filtering, and MongoDB CRUD operations built while learning modern Next.js.

Next.js 14App Router
NextAuth.jsGoogle OAuth
MongoDBMongoose CRUD
Next.jsReactMongoDBNextAuth.jsTailwind CSS
Case study
Source
08
Academic / CourseworkMachine Learning

Fake News Detector

University NLP classification project comparing Bidirectional LSTM architectures.

Cleans and tokenizes article text with NLTK and trains a Keras Bidirectional LSTM to identify fabricated news stories from linguistic patterns.

BiLSTMKeras / TF
NLTKPreprocessing
AcademicNLP Project
PythonTensorFlowBiLSTMNLTKScikit-learn
Case study
Source
09
Academic / CourseworkMachine Learning

Emotion Recognition

Transfer learning computer vision project fine-tuning ResNet50V2 on FER-2013.

Classifies facial expressions into 7 emotions using a fine-tuned ResNet50V2 model on the FER-2013 dataset.

FER-2013Benchmark Data
ResNet50V2Model Backbone
7 ClassesEmotions
PythonTensorFlowKerasResNet50V2OpenCV
Case study
10
Academic / CourseworkMachine Learning

Boston House Price Prediction

End-to-end regression pipeline comparing linear and tree models with Flask inference.

Analyzes 13 housing census attributes with Pandas, packages a trained Scikit-learn estimator via joblib, and serves predictions through a lightweight Flask web form.

Scikit-LearnRegression
joblibPipeline Export
Flask UIInference Form
PythonScikit-learnFlaskPandasNumPy
Case study
Source
11
Personal ProjectWeb & Full-Stack

Real-Time Chat App

Decoupled real-time messaging architecture using Next.js 15 and Socket.IO.

Connects a Next.js 15 client to an Express server over WebSockets, supporting scoped chat rooms, broadcast typing indicators, and user online status.

Socket.IOWebSockets
Next.js 15Client UI
MongoDBMessage History
Next.js 15React 19Socket.IOExpressTailwind CSS
Case study
Frontend Backend
12
Academic / CourseworkWeb & Full-Stack

Buxom Cosmetics

Multi-tier e-commerce architecture with Prisma ORM and MySQL schemas.

A university web project split across a React and Redux storefront and an Express REST API, modeling relational product variants, carts, and order checkout.

React + ReduxFrontend
Prisma + MySQLRelational DB
JWT AuthRoute Security
ReactTypeScriptExpressPrismaMySQL
Case study
Frontend Backend
13
Academic / CourseworkWeb & Full-Stack

POS Pharmacy System

Desktop retail pharmacy checkout system with FEFO batch inventory tracking.

A Java Swing and Hibernate 5.6 application for community pharmacies featuring barcode scanning, transaction rollbacks, and First-Expiring-First-Out medicine allocation.

Java SwingDesktop GUI
Hibernate 5.6ORM Persistence
FEFO LogicBatch Expiry
JavaJava SwingHibernate 5.6MySQL
Case study
Source
Browse all GitHub repositories

04 / Stack

The tools I use, from the interface to the infrastructure underneath it.

AI & data

LLMsRAGAgentic AIAnthropic ClaudeOpenAI Agents SDKOpenAI APICrewAILangGraphAutoGenMCPLiveKitPrompt EngineeringLangChainHugging FaceFine-tuningVector SearchVector Databases/StoresNLPTensorFlowKerasScikit-learnPandasNumPyMatplotlibSeabornOpenCVWhisperGradio

Backend & APIs

FastAPIDjango / Django REST FrameworkFlaskNode.jsExpressASP.NET CorePrisma ORMMongooseJWTSocket.ioSwagger / OpenAPIREST APIs

Frontend

Next.jsReactTailwind CSSRedux / Redux ToolkitReact NativeGSAPLenisFramer MotionHTMLCSSAndroid Studio

Platforms & storage

AWSRoute 53AWS ECSAWS EKSAWS S3AWS LambdaAWS API GatewayAWS ServerlessMicrosoft AzureGoogle Cloud PlatformVercelLinuxDockerKubernetesTerraformGrafanaPrometheusLokiPostgreSQLMySQLMS SQL ServerMongoDBAzure Cosmos DBFirebaseSupabaseUpstash Redis

Languages & delivery

PythonJavaScriptTypeScriptJavaC#C++SQLBash / ShellGit/GitHubCI/CDGitHub ActionsJenkinsGitLabRabbitMQ

05 / Services

The kinds of work I'm most useful for.

AI features

Adding LLMs, RAG, and agents to a product: reading and writing documents, answering questions from your own data, and automating routine steps.

Full-stack web apps

Websites and web apps from the interface down to the API and database, including login and role-based access. Mostly Next.js, React, Node.js, Python, and ASP.NET Core.

Platform engineering

Setting up the infrastructure apps run on: Docker, Kubernetes, Terraform, and monitoring with Grafana, Prometheus, and Loki.

Cloud and deployment

Deploying to Azure, AWS, or Vercel with CI/CD pipelines (GitHub Actions, Jenkins, GitLab CI), so a release is routine instead of a big event.

06 / Training

These are the courses where I built projects instead of just watching videos.

In Progress

Ed Donner / Ligency · Udemy

AI Engineer · Production Track

  • Deploying production-grade LLM applications, multi-agent architectures, and MCP across AWS (Bedrock, SageMaker, Lambda, SQS, Aurora), GCP, Azure, and Vercel.
  • Building 4 production products: SaaS Healthcare App (Next.js & App Runner), Digital Twin Mk II (Terraform & Bedrock), Cybersecurity Analyst Agent with MCP, and Multi-Agent SaaS Financial Planner with LangFuse observability.
AWS BedrockBedrock AgentCoreMCPAWS SageMakerTerraformLangFuseMulti-Agent
Course link (Certificate pending)

Ed Donner / Ligency · Udemy

AI Engineer · Agentic Track

  • Built eight AI agent projects using OpenAI Agents SDK, CrewAI, LangGraph, AutoGen, and MCP.
  • The final project was a simulated trading floor where four agents work together, make trades on their own, and use tools through MCP servers.
AgentsMCPCrewAILangGraphAutoGen
View certificate

Ed Donner / Ligency · Udemy

AI Engineer · Core Track

  • Built eight LLM apps in eight weeks, using Hugging Face, LangChain, RAG with vector search, and QLoRA fine-tuning.
  • Also built multi-agent systems and compared open-source and commercial models on coding and business tasks.
LLM engineeringRAGQLoRAFine-tuningHuggingFace
View certificate

Imran Teli · Udemy

Decoding DevOps

  • Learned DevOps through hands-on projects with AWS, Linux, Docker, Kubernetes, Terraform, Jenkins, GitHub Actions, and GitLab CI.
  • Also covered monitoring, and used GitHub Copilot and Amazon Q to help write scripts and automate things.
AWSGCPDockerKubernetesTerraformGitOpsObservability
View certificate

Charles Severance / University of Michigan · Coursera

Building Web Applications in Django

  • Learned how Django connects to databases using models, migrations, and ORM in SQLite, plus Python OOP and MVC patterns.
  • Built views with URL routing, template inheritance using the Django Template Language (DTL), and secure forms with CSRF protection.
DjangoPythonORMDatabasesFormsDTLMVC
View certificate

Charles Severance / University of Michigan · Coursera

Web Application Technologies and Django

  • Learned how dynamic web apps work: network sockets, the HTTP request-response cycle, HTML, CSS, and basic SQL.
  • Built a simple web server and browser in Python, set up a Django app, and deployed it live on PythonAnywhere.
DjangoPythonHTTPHTML & CSSSQLPythonAnywhere
View certificate

DeepLearning.AI · Coursera

Supervised Machine Learning

  • Learned to build and train regression and classification models in Python with NumPy and scikit-learn.
  • Covered linear and logistic regression, models with several input features, and regularization.
Machine learningRegressionClassificationPython
View certificate

Academind by Maximilian Schwarzmüller · Udemy

React · The Complete Guide

  • Learned React components, hooks, forms, routing, Context API, and Redux Toolkit.
  • Also covered an intro to Next.js and how to deploy React apps.
ReactNext.jsReduxFrontend
View certificate
What these courses coveredAgentic AI · RAG · LLMs · CI/CD · Cloud infrastructure

07 / Leadership & Volunteering

What I did at university besides studying: sponsorship work, event teams, and volunteering.

Leadership

Dec 2023 — Mar 2024

Aspire Leaders Program

Aspire Institute · Remote

  • Completed 30 hours of coursework, including three full modules and a culminating project.
  • Final project: "Empowering Minds - Education Outreach for Needy Children in Pakistan", a plan to help children in need get a better education.
LeadershipCommunity workEducation outreachTeamwork
View certificate
Aug 2022 — May 2023

Deputy Head of Marketing

SOFTEC'23 · FAST-NUCES, Lahore, Pakistan

  • Helped lead a marketing team of about 40 people, before and during the event.
  • Worked with company executives to close three sponsorship deals. We raised over PKR 1,000,000, which was 25% above the target.
Team leadershipSponsorshipsPartnershipsMarketing
View certificate

Volunteering

Nov 2022 — Jan 2023

Volunteer · Operations

Future Fest'23 · Lahore, Pakistan

  • Checked billboards, banners, and other promotional materials.
  • Looked after security at the auditorium and VIP areas while working with senior police officers, including the District Police Officer.
  • Helped set up and pack up the event, and stayed in touch with teams and vendors.
Event operationsLogisticsSecurity
View certificate
Oct 2021 — Aug 2022

Volunteer · Marketing, Software House Enclosure & Infrastructure

SOFTEC'22 · FAST-NUCES, Lahore, Pakistan

  • Called HR managers and CEOs to set up sponsorship meetings, and joined those meetings with the marketing head.
  • Worked with the setup team to run the Software House Enclosure.
  • Looked after company exhibits and talked with visitors to keep things running smoothly.
MarketingSponsorshipsSponsor outreachEvent management
View certificate

08 / Contact

Working on something? Let's talk.

Tell me what you're working on. I'm open to full-time roles in Lahore or remote, and happy to talk about AI, backend work, or web apps. Email is the quickest way to reach me.

Book a callEmail me