I worked as a Software Engineer & App Developer at Auburn University's Startup Studio, building MVPs and scalable prototypes for student-led ventures using modern web technologies, from "idea, building to growth".
We build & ship enterprise applications in 30-40 days. Crossed $120,000 in all-time agency revenue in 8 months. Helping clients generate $60,000+ in revenue overall, 3,000+ signups, and attract nearly $100K in angel investment overall (e.g one of the client brandjet.ai)
A B2B lead-gen platform that delivers 40% more verified niche leads via AI-powered filtering across LinkedIn, Instagram & TikTok. Scaled to $40K+ in revenue in 3 months with 100% inbox rate and 80% open rates through domain warmups, 3x lead vetting, and precision cold email automation.
AI-powered User-Generated Content (UGC) video creation platform enabling brands to generate high-quality video ads in minutes instead of days. Built with 100+ AI avatars, multi-language support (29 languages), and automated script generation.
Built an AI writing/research tool that generates essays/outlines/citations ~3× faster, with real-time grammar, structured outlining, APA/MLA/Chicago formatting, and a Homework Helper to improve productivity and accuracy.
Premium Next.js SaaS starter kit enabling developers to launch full-featured apps. Including Auth, Payments, AI, and SEO, in hours, not weeks. Generated $30,000 in revenue within just 2 months, outpacing many competitors by delivering a 10× faster build-to-launch workflow. Included in ProductHunt top 7 performers of the day.
AI-powered SEO platform for analysis, research, and content generation, scaled to $50,000 revenue and 6,000 users in 3 months by collaborating with 10+ SEO experts to craft high-performing backend prompts and most efficient auto-ranking outputs. Included in ProductHunt top 5 performers of the day.
Enabling businesses to create customizable AI chatbots for 24/7 automated support, resolving up to 70% of incoming inquiries instantly, reducing ticket volume, and freeing human agents for complex cases.
Securely embedded on company's data (FAQs, Docs, Pages, etc) to get instant, updated and efficient answers to the user's queries.
AI-powered chat engine for interactive document conversations, leveraging vector databases and per-user agents with large context windows, integrated with a PDF viewer for real-time section highlighting. Included in ProductHunt top 10 performers of the day.
An AI-powered avatar and profile‑picture generator that produces 300+ unique, "crispy" avatars in a single batch, one single API for all avatar generation needs. Launched before Lensa came hype came into the market. Grown to 1000 users and $3000+ in 1 week. Included in ProductHunt top 7 performers of the day.
Buy & Sell Early-Stage Startups with early traction, opportunity for investors to get ready made businesses. Included in ProductHunt top 5 performers of the day.
Resource marketplace for photographers, get hundreds of presets in just one click.
Web3 Defi App PastrySwap for midcap coins. Providing ease to buy, sell, and swap coins.
AI-powered platform that enables users to create, host, and distribute professional voiceovers in seconds, leveraging 400+ natural-sounding voices to accelerate content production compared to manual recording, for several different user personas (e.g voice for blogs, voice for e-newspaper, voice for youtube videos, voice for podcasts)
Huge bundle of resources related to untapped/under-represented websites with outstanding features & access to special offers only for untapped members, along with 10,000+ automated guest posting opportunities.
LACOS is a cosmetics startup that produces organic & natural skincare products.
AI‑powered student search engine that provides instant access to 400,000+ A‑Level and O‑Level exam solutions from all the exam zones, with auto fetching youtube videos related to questions and answers, cutting educational cost by 90% across South and Southeast Asia.
Launched a content marketplace connecting students, parents, and teachers to buy and sell student-written resources, expanding educational access and affordability in targeted regions.
Provides transformational digital technology solutions and services.
One of the biggest AdTech startups in Pakistan. Grown to 2 million per month in website traffic and a network of 1000+ traffic providers, brokering the traffic from providers to big adtech companies (Adsense, Revcontent, Taboola)
Master of Computer Science & Software Engineering
Auburn, AL | 2025 - Ongoing
Bachelor of Computer Science
Peshawar, Pakistan | 2020-2024
Alevels
Gujranwala, Pakistan | 2018-2020
Olevels
Gujranwala, Pakistan | 2016-2018
Despite remarkable progress in large language models, Urdu—a language spoken by over 230 million people—remains critically underrepresented in modern NLP systems. We introduce Qalb, an Urdu language model developed through a two-stage approach: continued pre-training followed by supervised fine-tuning. Starting from LLaMA 3.1 8B, we perform continued pre-training on a dataset of 1.97 billion tokens, achieving state-of-the-art performance with a weighted average score of 90.34, outperforming the previous best model by 3.24 points and demonstrating a 44.64-point improvement over the base model.
ScopeTags is a retrieval-augmented generation framework using three token types, retrieval gating ([Ret]/[NoRet]), relevance scoring ([Rel]/[NoRel]), and consistency filtering ([Cst]). To enable multimodal models to autonomously decide when to retrieve external knowledge, filter irrelevant documents, and resolve contradictions. It achieves up to +8.2% improvements on knowledge-intensive VQA benchmarks by leveraging the model's self-knowledge as a governance signal for knowledge integration, eliminating the need for external annotations.
SLM-first (Small Language Model) agent architectures over LLM-centric designs, showing that compact models (<10B parameters) achieve superior latency (45ms vs 320ms), cost efficiency (10-30x cheaper), and schema compliance (94% vs 87%) for specialized agent tasks while maintaining comparable performance. It provides a 6-phase migration framework and identifies adoption barriers, demonstrating that SLMs enable economically sustainable, edge-deployable agent systems for enterprise applications.
Built a Flutter/Firebase app that generates personalized itineraries via chat, recommending flights/hotels/activities with booking links; powered by a CrewAI + LangChain multi-agent planner, Falcon LLM, Pinecone semantic search, and Flask APIs aggregating real-time data from 30+ travel sources (Airbnb/Booking/TripAdvisor).