hi, saja here

Experienced Software Engineer with a strong background in designing, developing, and maintaining full-stack and AI/LLM systems. I thrive on demystifying ambiguous problems, collaborating across diverse teams, and driving complex projects to completion.

TypeScriptJavaPythonAngular

In my free time, I'm nerdy about tech and dive into psychology books.

Recently Careem, an Uber Company

Work

  1. Software Engineer II, AI Systems

    May 2025 – Dec 2025

    Careem

    • Architected a multi-agent LLM orchestration system using LangGraph and Model Context Protocol, enabling automated workflow generation through GitHub and Jira integrations.
    • Designed and built an NLP-to-SQL translation system using RAG and vector similarity search, improving semantic and logical query accuracy for business users.
    • Built a systematic AI evaluation framework using LangSmith and LLM-as-Judge methodology to validate model outputs against organizational quality standards.
    • Designed and presented an AI automation proof-of-concept to executive leadership, securing approval and budget for production rollout.
    PythonLangChainLangGraphLangSmithModel Context ProtocolRAGVector Similarity SearchLLM-as-JudgeGitHub CopilotCursorAzure
  2. Software Developer, Full Stack

    Jun 2022 – May 2025

    Constellation Homebuilder Systems

    • Designed and architected a full-stack vendor invoice submission module, providing a unified interface for external vendors across consolidated ERP systems.
    • Built complex wizard-based workflows with reactive forms, client-side validation, routing logic, and REST API integrations to support multi-step submission processes.
    • Built and maintained backend services and APIs using .NET Framework and ASP.NET, integrated with SQL Server and PostgreSQL, supporting core application logic.
    • Optimized application performance through database query tuning, frontend rendering improvements, and API response time reductions.
    AngularTypeScriptRxJS.NET FrameworkASP.NETSQL ServerPostgreSQLREST APIs

Reviews

The way Saja designed and delivered our AI engineering agent, and effectively demonstrated its capabilities to the broader organization, is a testament to this rare combination of strengths. She has significant potential, and this achievement marks only the beginning of what I believe will be an outstanding journey.

Khurram Naseem

Director at Careem · 25+ years in the industry

linkedin ↗

I worked with Saja on our most forward-thinking AI initiatives to improve our SDLC. She is a strong engineer and capable of presenting to executive-level colleagues.

Jonathan Keenan

Director at Careem

linkedin ↗

Tech Stack

// Languages

PythonGoJavaC#SQLJavaScriptTypeScript

// Backend & APIs

.NET Framework 4.7+ASP.NET CoreREST APIsGraphQLMicroservicesEvent-Driven Architecture

// AI / LLM Engineering

LangChainLangGraphLangSmithRAGVector Similarity SearchLLM-as-JudgeNLP-to-SQLModel Context Protocol (MCP)Multi-Agent OrchestrationPrompt Engineering

// Frontend

AngularTypeScriptRxJSReactJavaScriptHTML5CSS3

// Databases & Caching

SQL ServerPostgreSQLRedis

// Cloud & DevOps

AzureAzure FunctionsAzure Service BusAWS (working knowledge)DockerKubernetesCI/CD PipelinesJenkins

Case Studies

Deep dives into production-scale engineering problems, where the challenge wasn't just writing code, but shipping safely across distributed systems, multiple teams, and real users.

Careem Consumer App

Multi-Region Currency Symbol Migration

Production rollout · Distributed systems · Multi-team coordination · Zero-downtime constraint

25+

Repos in a production codebase touched

10+

Squads whose release cycles I coordinated

Multi-region

Live markets affected simultaneously

0

Customer-facing incidents despite the scope

$ cat problem.md, approach.md, outcome.md

This was a production change to a live, high-traffic app used across multiple regions and currencies, the kind of change where a mistake shows up directly in what customers see when they're about to pay. It touched 25+ repositories, including legacy code that predated the current architecture, across nearly every surface in the consumer app where a price or fare appeared.

I implemented and helped drive the rollout across Careem's iOS and Android consumer apps, coordinating with 10+ engineering squads across multiple regions to make sure the change landed consistently everywhere pricing was shown, rather than shipping it in isolation and hoping it propagated cleanly.

Working across legacy and current code added real complexity: older modules often handled currency formatting differently or lacked the conventions the newer codebase relied on, so validating consistent behavior meant testing each surface individually rather than assuming one fix would apply everywhere.

  • Validated the change on internal builds first, across both iOS and Android, to catch platform-specific and legacy-code issues early.
  • Tracked results as we expanded to Careem employees company-wide, a much larger and more varied internal test group.
  • Moved to a gradual customer rollout only after both internal phases cleared, watching each region before expanding further.
  • Stayed in sync with the QA manager at each phase transition so sign-off was a deliberate gate, not an assumption.

Shipped a production-wide change across 25+ repositories and 10+ squads spanning multiple live markets, with zero customer-facing incidents, validated through a deliberate, checkpoint-gated rollout rather than a single big-bang release.

Projects

This is where I experiment with new ideas, technologies, and concepts.

resume-jd-matcher.py

Resume–JD Semantic Matcher

Embeddings-based match scoring between a resume and a job description, with LLM-generated gap analysis. Three-section UI: resume upload, JD link input, matching/missing concepts results.

PythonFastAPIReactsentence-transformersOllamaembeddings
links coming soon
cold-email-generator.py

GenAI Cold Email Generator

Point it at a company's careers page and it extracts the job listings, then generates a personalized cold email for each, pulling relevant portfolio links from a vector database matched to the specific job description, so the pitch actually fits the role.

PythonLangChainLlama 3.1 (Groq)ChromaDBStreamlit
links coming soon
meeting-scheduler.py in progress

Phased Meeting Scheduler

A scheduling tool that resolves availability across multiple participants in phases rather than a single round-trip, narrowing down time windows progressively instead of requiring everyone to respond at once.

links coming soon

// 06 — get in touch

let's talk_

Open to new opportunities and interesting problems.

sajakh00@gmail.com linkedin ↗ github ↗ resume ↓

// built and designed by Saja Khan

© 2026 Saja Khan. All rights reserved.