Skip to content
View arushahmd's full-sized avatar
  • Freelance
  • Vaughan, Ontario, Canada | Open to Canada + Remote
  • 18:56 (UTC -12:00)

Highlights

  • Pro

Block or report arushahmd

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
arushahmd/README.md

Aroosh Ahmad

Applied AI Engineer · Production AI Systems · LLM/NLP · Voice AI · Computer Vision · Python Backends

Portfolio LinkedIn Email

I build production-oriented AI systems across LLM/NLP, real-time voice, computer vision, and Python backend infrastructure.

My engineering approach is to use AI where probabilistic understanding is valuable, while keeping orchestration, validation, state, and side effects explicit and reliable. My work sits at the intersection of applied AI, software engineering, and experimental research.

Selected Work

Real-time conversational ordering system combining Twilio Media Streams, Deepgram STT/TTS, custom NLU, Redis-backed sessions, and deterministic state-machine orchestration.

Signal: real-time AI · conversational systems · backend architecture · state management · observability · deployment

Transformer-based hierarchical intent classifier with independent main/sub-intent heads, leakage-resistant splitting, reproducible experiment manifests, evaluation, offline tests, and CI.

Signal: NLP · transformers · evaluation · reproducibility · ML engineering

Reproducible classical computer-vision pipeline for recovering degraded DataMatrix symbols with explicit inference rejection and deterministic synthetic evaluation.

  • Exact decode improved from 28/35 to 33/35 on the committed synthetic benchmark
  • Grid shape was exact for 34/34 accepted cases

Signal: computer vision · image processing · benchmarking · failure handling · reproducibility

Empirical study of mini-batch composition and training-example order for instruction fine-tuning across 54 reviewed runs using Flan-T5, LoRA, FAISS, and multi-seed evaluation.

Signal: LLM fine-tuning · experiment design · multi-seed evaluation · research methodology

Core Stack

Core stack

AI / LLM: LLMs, RAG, Agentic AI, tool workflows, prompt engineering, fine-tuning, evaluation, PyTorch, Transformers, Hugging Face, LoRA/PEFT, FAISS
Backend & Data: Python, FastAPI, REST APIs, WebSockets, SQL, PostgreSQL, Redis
Voice & Vision: Twilio, Deepgram, streaming STT/TTS, OpenCV, OCR, image processing
Infrastructure: Docker, GitHub Actions, AWS, Azure

Background

  • AI / Backend Engineer — Cygnus Payments
  • Independent LLM Research
  • AI Research Officer — Center of Language Engineering
  • AI Engineer — Nodlays
  • MPhil Artificial Intelligence — PUCIT
  • BS Computer Science — Lahore Garrison University

Connect

Portfolio · LinkedIn · Email

Pinned Loading

  1. restaurant-voice-ai restaurant-voice-ai Public

    Real-time AI voice ordering system using FastAPI, Twilio Media Streams, Deepgram STT/TTS, Redis, custom NLU, and deterministic state-machine orchestration.

    Python 1

  2. llm-batching-research llm-batching-research Public

    Empirical study of semantic mini-batch construction and length-based curriculum ordering for instruction fine-tuning using Flan-T5, LoRA, FAISS, and multi-seed evaluation.

    Jupyter Notebook

  3. urdu-document-ocr urdu-document-ocr Public

    End-to-end Urdu document OCR with RTL layout analysis, PyTorch CNN-BiLSTM-CTC training and inference, reproducible synthetic evaluation, CLI, and FastAPI.

    Python

  4. multihead-intent-classification multihead-intent-classification Public

    Multi-head transformer intent classifier with reproducible training, evaluation, experiment tracking, and historical error analysis.

    Python

  5. datamatrix-reconstruction datamatrix-reconstruction Public

    Classical computer-vision pipeline for reconstructing degraded DataMatrix symbols with deterministic synthetic evaluation and finder-pattern grid inference.

    Python