Skip to main content

Open to AI & backend engineering roles · Karachi, Pakistan

I build retrieval systems and the backends that serve them.

Computer Science student at KSBL, currently an AI engineering intern at the State Bank of Pakistan — building RAG pipelines, serving LLMs with vLLM, and writing the FastAPI services in between.

draw a digit — a small neural net I trained reads it, layer by layer:

mlp 196→32→16→1096.7% MNIST test acc · in-browser
prediction: 7 · 99.9% confidentThe network predicts 7

What's actually in the stack

Infrastructure, models, frameworks, and database technologies used to build scalable AI microservices.

LLMs & Foundation Models

Qwen3 (72B/14B)Core
Primary reasoning model for document extraction & RAG
Llama 3 / 3.1Core
Agentic tool calling & multi-turn dialog execution
DeepSeek R1 / V3
Structured JSON generation and math verification

Frameworks & Agent Orchestration

LangGraphCore
Stateful multi-agent workflows with tool execution loops
LangChainCore
Document processing, text splitters & retrieval chains
RAGAS Framework
Production evaluation metrics for hallucination & faithfulness

Retrieval & Vector Storage

pgvector (PostgreSQL)Core
Production vector indexing with HNSW cosine search
BAAI bge-reranker-largeCore
Two-stage cross-encoder re-ranking pipeline
Apache Tika
Multi-format PDF, DOCX, and regulatory document extraction

Model Serving & Inference

vLLMCore
High-throughput PagedAttention LLM inference server
OllamaCore
Local LLM serving & quantization testing harness
Open WebUI
Enterprise UI interface & role-based access client

Backend & Databases

FastAPI (Async Python)Core
High-performance microservices & SSE streaming endpoints
PostgreSQL
Relational data store & JSONB document schema
Redis
Distributed response caching & pub/sub message queue

DevOps & Observability

Docker & ComposeCore
Multi-container orchestration for 5+ microservices
Microsoft Azure
Cloud VM instances, storage buckets & networking
Prometheus & Grafana
Real-time latency, GPU metrics & service health dashboards

About

I'm a Computer Science student at KSBL, currently interning as an AI engineer at the State Bank of Pakistan. Most of my time goes into retrieval-augmented generation — ingestion, retrieval, reranking — and the backend work that keeps it running: FastAPI services, LLM serving, and enough observability to know when something's actually broken.

Enterprise RAG
Ingesting 3,900+ regulatory documents with Apache Tika, pgvector HNSW indexing, and a BAAI cross-encoder reranking stage.
LLM serving
Serving Qwen3 and Llama with vLLM (PagedAttention) and Ollama on Azure GPU instances, streamed over SSE.
FastAPI backends
Async REST endpoints, Server-Sent Events, Pydantic schemas, and Redis caching.
Observability
Latency, GPU memory, and RAGAS hallucination scores tracked in Prometheus and Grafana.

on the résumé

  • Enterprise AI Engineer Intern @ State Bank of Pakistan
  • Architected AskSBP Assistant indexing 3,900+ documents
  • Deployed Qwen3 models via vLLM with pgvector hybrid search
  • Achieved 40-50% hallucination reduction via RAGAS testing
  • Built 5+ FastAPI microservices containerized with Docker
  • Integrated Prometheus & Grafana observability stack
  • Engineered Smart Firing Range Computer Vision solution
  • Co-founded KSBL ACM Student Chapter & Tech Carnival
  • Teaching Assistant for Data Analytics & Programming

Projects

Detailed system architecture flows, RAG pipelines, multi-agent frameworks, and measurable outcomes.

ProductionEnterpriseRAGLLMAI InfrastructureBackendAzureDockerized
3,900+ Docs

AskSBP Enterprise AI Assistant

State Bank of Pakistan Regulatory RAG Platform

Production-grade internal enterprise AI assistant built for the State Bank of Pakistan to ingest, search, and synthesize insights across 3,900+ banking circulars, compliance policy manuals, and financial regulations.

System Architecture Flow
Regulatory PDF / DOCX
Apache Tika Parsing
Sentence Chunking
pgvector (HNSW Index)
BAAI Cross-Encoder Rerank
Qwen3 via vLLM
Open WebUI / SSE Stream
FastAPIQwen3vLLMOllamapgvectorBAAI RerankerApache TikaOpen WebUIDocker ComposeAzurePrometheusGrafanaRAGAS
EnterpriseAgentic AILLMBackendPythonFastAPI
Multi-Agent State Machine

Financial Multi-Agent Research Platform

Autonomous Market Research & Financial Synthesis

Autonomous multi-agent framework built with LangGraph and LangChain that orchestrates specialized research agents for financial market analysis, balance sheet extraction, news sentiment synthesis, and automated report generation.

System Architecture Flow
User Research Request
Router Agent (LangGraph)
Data Retrieval Tool
Financial Calculator Tool
Synthesis Agent
Structured Report PDF
LangChainLangGraphFastAPIPostgreSQLRedisPythonDocker
ProductionBackendFastAPIPythonMachine Learning
Macro-Economic Forecasts

SBP Economic Forecasting Dashboard

Time-Series Economic Indicators & Analytics Engine

Full-stack economic indicator analytics platform integrating State Bank of Pakistan financial APIs with statistical and machine learning forecasting models (Prophet, ARIMA, XGBoost) for macro-economic trend analysis.

System Architecture Flow
SBP Macro API
FastAPI Data Ingestion
Feature Engineering
Prophet / ARIMA Models
PostgreSQL Cache
React Interactive Dashboard
FastAPIReactPythonProphetARIMAXGBoostPlotlyPostgreSQL
ProductionComputer VisionMachine LearningPython
Sub-Millimeter Shot Detection

Smart Firing Range AI System

Computer Vision Target Detection & Ballistic Telemetry

Mission-critical defense AI system developed at Milita Tech Solutions incorporating real-time Computer Vision (OpenCV) for automated target detection, sub-millimeter shot coordinate analysis, and ballistic accuracy telemetry.

System Architecture Flow
High-Speed Camera
OpenCV Frame Capture
Target Mask Segmentation
Shot Coordinate Extraction
Ballistic Analytics
PythonOpenCVMachine LearningComputer VisionEmbedded SystemsC++
Milita Tech Defense AI
ProductionAI InfrastructureDockerizedAzureMonitoring
Zero-Downtime Containers

AI Infrastructure & Deployment Platform

Containerized LLM Orchestration & Monitoring Stack

Production infrastructure setup featuring containerized LLM inference nodes, Nginx load balancing, Redis response caching, and full observability powered by Prometheus metrics and Grafana dashboards.

System Architecture Flow
Nginx Reverse Proxy
FastAPI Load Balancer
vLLM Inference Pool
Redis Response Cache
Prometheus Scraper
Grafana Dashboard
DockerDocker ComposeRedisPostgreSQLNginxPrometheusGrafanaAzure
Milita Tech Defense AI

Skills

Categorized technologies utilized in enterprise AI systems, high-throughput microservices, and cloud deployments.

AI Engineering

Generative AI, Large Language Models, RAG Pipelines, Multi-Agent Orchestration

LLMs (Qwen3, Llama)
Enterprise RAG
Agentic AI
LangChain
LangGraph
Prompt Engineering
RAGAS Evaluation
HuggingFace Transformers
Sentence Embeddings
pgvector
Hybrid Retrieval

Experience

Roles, central bank RAG deployments, defense AI systems, and academic teaching assistantships.

🏦

Enterprise AI Engineering Intern (ITPMD)

State Bank of Pakistan (SBP)
Jun 2026 – PresentKarachi, Pakistan

Architecting enterprise AI systems and internal RAG applications supporting digital transformation initiatives at Pakistan central bank.

Key Outcomes & Engineering Deliverables
Architected AskSBP Enterprise AI Assistant indexing 3,900+ central bank circulars & regulatory policy documents.
Deployed Qwen3 LLMs using vLLM PagedAttention and Ollama serving engines hosted on Azure VMs.
Engineered two-stage hybrid retrieval combining pgvector (HNSW) vector search with BAAI bge-reranker-large cross-encoders.
Delivered 5+ asynchronous FastAPI microservices with SSE streaming response endpoints.
Achieved 40-50% reduction in RAG hallucination rates verified via RAGAS framework evaluations.
Configured full containerized infrastructure using Docker Compose, monitored with Prometheus and Grafana dashboards.
FastAPIQwen3vLLMOllamapgvectorBAAI RerankerApache TikaOpen WebUIAzureDocker ComposePrometheusGrafanaRAGAS
⚙️

Junior Systems & AI Engineer

Milita Tech Solutions
Feb 2026 – Jul 2026Karachi, Pakistan

Engineered machine learning and computer vision solutions for defense and military system integration.

Key Outcomes & Engineering Deliverables
Engineered Smart Firing Range solution using Python, OpenCV, and Machine Learning for real-time target detection.
Developed computer vision algorithms for sub-millimeter shot placement coordinate extraction and ballistic analytics.
Contributed to secure frequency-hopping communication protocols for Unmanned Aerial Vehicle (UAV) control links.
Integrated embedded systems hardware with backend processing pipelines for mission-critical applications.
PythonMachine LearningComputer VisionOpenCVTarget DetectionShot AnalysisFrequency HoppingEmbedded Systems
🎓

Teaching Assistant (Data Analytics & Programming)

KSBL
Oct 2025 – Feb 2026Karachi, Pakistan

Mentored Computer Science and Business Analytics students in Python, Data Analysis, and Machine Learning.

Key Outcomes & Engineering Deliverables
Mentored 60+ students in Python programming, pandas data pipelines, and machine learning fundamentals.
Guided lab practicals covering SQL query optimization, data cleaning, and statistical modeling.
Collaborated with course faculty to review coding assignments, projects, and grading rubrics.
PythonMachine LearningData AnalysisPandasNumPySQL
🚀

Co-founder & Chief Operating Officer

KSBL ACM Student Chapter
Oct 2024 – Jun 2026Karachi, Pakistan

Co-founded university tech community chapter and led flagship AI conferences & technical workshops.

Key Outcomes & Engineering Deliverables
Planned and executed "AI Unleashed — The Tech Carnival", attracting 300+ attendees and industry leaders.
Organized technical workshops on Generative AI, Python backend development, and robotics.
LeadershipOperationsAI CommunityEvent Planning

Timeline

Evolution from Computer Science foundation at KSBL to Enterprise AI & RAG Engineering.

1
2023 – PresentBSCS Candidate

BS Computer Science Student

Karachi School of Business and Leadership (KSBL)

Specializing in Enterprise Software Engineering, Artificial Intelligence, Database Systems, and Distributed Infrastructure.

Started BSCS degree
Maintained top academic standing
Deep focus on Python, Systems Architecture, and Databases
2
2025Data Analytics TA

Teaching Assistant

KSBL Department of Computer Science

Supported delivery of Data Analytics and Programming coursework for undergraduate students.

Mentored 60+ students in Python & ML
Led lab assignments & code reviews
3
2025Co-Founder & COO

Co-founded ACM Student Chapter

KSBL ACM Chapter

Established the official ACM student chapter to build an active technology and AI community.

Co-founded ACM Chapter as COO
Organized "AI Unleashed" Tech Carnival
4
2026CV & Systems Engineer

Systems & AI Engineer

Milita Tech Solutions

Developed computer vision models for target detection and secure communication protocols for UAV systems.

Built Smart Firing Range CV system
Implemented shot detection algorithms
Worked on Frequency-Hopping UAV links
5
2026 – PresentEnterprise AI Engineer

Enterprise AI Engineering Intern

State Bank of Pakistan (SBP)

Building production enterprise RAG systems, deploying vLLM inference microservices, and setting up AI observability.

Built AskSBP indexing 3,900+ docs
Deployed 5+ FastAPI services
Reduced hallucinations by 40-50% via RAGAS
6
PresentProduction AI Specialty

Building Enterprise AI Systems

Production AI & Cloud Infrastructure

Architecting scalable AI infrastructure, RAG pipelines, multi-agent frameworks, and high-concurrency backend services.

Available for Enterprise AI & Backend Roles
Focus on LLMs, RAG, Agentic AI, FastAPI

Certifications

Credentials from Oxford/Harvard's Legatum Institute, DataCamp, and Coursera.

Valid until 2033

Leading with Character

University of Oxford (Legatum Institute & Harvard)

Issued Mar 2025
LeadershipStrategic Decision MakingEthical Governance
View Credential

Professional SQL & Database Querying

DataCamp

Issued Dec 2024
PostgreSQLComplex JoinsQuery OptimizationData Aggregation
View Credential

Google Data Science Foundations: Data, Data, Everywhere

Coursera (Google)

Issued May 2024
Data ScienceData Analysis PipelinesData VisualizationSQL
View Credential

Writing

Notes on retrieval, model serving, and what building AskSBP actually involved. Long-form, no gimmicks.

Open for Enterprise AI & Backend Roles

Looking for an AI Engineer?

Let's build production AI systems, high-throughput RAG pipelines, and resilient backend microservices together.