aleena khawaja

aleena khawaja

Software Engineer & AI Researcher specializing in Multi-Agent Systems and Distributed Computing

Based in: Lahore, PakistanMost recently: Teaching Assistant, CS 202 Data Structures, LUMSExperience: 1 year

Hire me

A Computer Science student at LUMS with a strong background in building autonomous drone swarms, distributed systems, and scalable ML pipelines. She has professional experience in LLM orchestration and full-stack development using technologies like Python, Go, and Docker.

Experience

Teaching Assistant, CS 202 Data Structures

LUMSLahore, Pakistanpart timeSep 2025 – Dec 2025
Held biweekly office hours to guide students through assignments, debugging, and core data structure concepts. Recorded tutorial sessions used as reference material by the broader student cohort across the semester.
Data Structures

Freelance Software Engineer

AI Storytelling Mobile AppfreelanceJul 2025 – Aug 2025
Developed a Python backend with scalable speech-to-text and AI story generation APIs. Integrated a Flutter frontend to deliver real-time interactive story generation for children. Integrated Supabase for authentication and persistent storage, supporting both guest sessions and secure account-based flows.
PythonFlutterSupabaseSpeech-to-textAI story generation APIs

Software Engineering Intern

RevsmithinternshipAug 2024 – Feb 2025

Summary

Developed the personalization module for a multi-agent LLM pipeline, iterating on prompt strategies to generate tailored outbound messages from structured company profiles. Built data ingestion modules using Firecrawl and LangChain to scrape, clean, and enrich unstructured web sources into structured schemas for downstream AI processing. Prototyped chain-of-thought and few-shot pipelines with DSPy and OpenAI APIs; designed custom metrics to programmatically evaluate LLM outputs for tone, relevance, and proof-pain alignment.

What I did

  • Prototyped prompt strategies and measured them offline using chain-of-thought and few-shot approaches.
  • Designed custom metrics to score generated emails on tone, relevance, pain-point alignment, and personalization.
  • Focused on inference and orchestration for the ML pipeline.
  • Built evaluation setups to measure how a model changes after an update.
LLMFirecrawlLangChainDSPyOpenAI APIsChain-of-thoughtFew-shotOpenAI API

Skills

Technical

VS Code
GitGit
PythonPython
MongoDBMongoDB
GolangGolang
FlaskFlask
LangChain
PyTorchPyTorch
DockerDocker
DSPy
scikit-learn
Whisper
pandas
NumPy
PostgreSQLPostgreSQL
ReactReact
BashBash
LinuxLinux
C++C++
FlutterFlutter
TypeScriptTypeScript
Node.jsNode.js
Mirascope
Speaker Diarization
Audio Segmentation
OpenAI API
AirSim
GAT actor-critic model
Model Training
Socket.io

Projects

Hierarchical Multi-Agent Exploration Framework

Senior Year Project Leadacademic
Built an end-to-end autonomous drone swarm system integrating cooperative exploration, formation control, YOLOv8 adversary detection, and pursuit-evasion, validated in AirSim high-fidelity simulation. Designed a GAT actor-critic with geographical K-Means clustering and risk-aware Value Iteration under a Centralized Training, Decentralized Execution (CTDE) paradigm for decentralized multi-agent coordination. Achieved 73% map coverage and zero inter-agent collisions with 26% fewer parameters and ∼3× faster inference versus centralized baseline, within a 2-3% coverage gap. Paper submitted to ICRA 2027 based on this work.
PyTorchAirSimYOLOv8GAT actor-criticK-MeansCTDE

Urdu Transcription Agent

personal
Extended Whisper's pipeline with custom segmentation logic and speaker labeling for long-form Urdu audio. Built a CLI tool producing punctuated, RTL-correct transcriptions using python-bidi and arabic-reshaper.
PythonWhisperNLPpython-bidiarabic-reshaper

Raft Consensus Algorithm Implementation

academic
Implemented Raft's leader election, log replication, and consensus protocol from the extended Raft paper, including RPC-based coordination between peers with no shared state. Built fault tolerance into the system with persistent storage, allowing servers to recover committed log state after a crash and restart. Validated correctness under leader failures, network partitions, and concurrent client requests using the official test suite, including Go's race detector.
GoDistributed SystemsRPC

ColorGrid

personal
A multiplayer game where players compete to claim the largest connected block of cells. It features a real-time layer for game state synchronization and matchmaking.
TypeScriptReactNode.jsExpressSocket.ioMongoDB

Note-Taking Web App with ML Inference Pipeline

personal
Built a full-stack app with JWT auth, Redis caching, and MongoDB, with an ML backend supporting summarization, quiz generation, auto-tagging, and text-to-speech. Decoupled Flask ML inference service from the core app using Docker to allow independent scaling of AI workloads.
FlaskMongoDBRedisDockerJWTReact

Senior Year Project

personal
A project completed during senior year that involved model training experience.

Education

B.Sc. Computer Science

Lahore University of Management Sciences (LUMS)Computer ScienceGPA 3.51Aug 2022 – May 2026

Coursework

Software EngineeringDatabasesData StructuresAlgorithmsMachine LearningAdvanced ProgrammingDistributed SystemsNetwork-Centric ProgrammingNetwork SecurityOperating SystemsDynamic Programming & Reinforcement Learning

Recognition

Awards

Power Pitch Award

Conrad Challenge2021

Runners-Up

Asian Regional Space Settlement Design Competition2021

100% Merit Scholarship, A-Level

A-Level Institution

Patents & publications

Hierarchical Multi-Agent Exploration Framework

publication2027

Paper submitted to ICRA 2027 based on Senior Year Project work.

Hire aleenaReach out about a role, a contract or a conversation.For recruiters

aleena's twin is AI, it can make mistakes.