Aryan — engineer

┌─┐┬─┐┬ ┬┌─┐┌┐┌
├─┤├┬┘└┬┘├─┤│││
┴ ┴┴└─ ┴ ┴ ┴┘└┘

$ whoami

I'm a 25 year old software engineer. Currently I work as a backend engineer specialising in ML/AI platforms. I enjoy building reliable systems that work fast, whether thats distributed systems, RAG pipelines, agentic systems or low-level systems straddling the hardware/software interface.

loc: Chicago, IL
job: Software Engineer (AI/ML Platform), Certara AI
edu: MEng Electronic & Information Engineering, Imperial College London
lang: en / fa

$ cat about.txt

My work sits somewhere between systems, performance, backend and AI/ML infrastructure: model serving, retrieval, distributed workflows, hot paths and the unpleasant plumbing between them. I get most of my enjoyment from fighting fires on-call, ghost-hunting through labyrinthine codebases and being the point guy when something needs to become correct, fast and boring. A pondering code monkey, but a code monkey at heart.

Lately that has meant building private agentic systems for large F100 clients and federal entities (contract). Words like 'agents', 'multimodal' and 'RAG' get thrown around like chewed-up dog toys; my job is to turn them into systems that are private, observable and boring enough to trust.

Outside work I mess with open-source development (vLLM and co), faffing around with my Flipper Zero, sharpening my low-level C/C++, low-latency systems, market structure, quantitative research, competitive programming and DeFi primitives.

$ git log --oneline work/

a4f19c

Software Engineer, Machine Learning Infrastructure — Certara AI (US)

2023—present · privatised LLM serving and agent building for pharma and federal clients

7b02de

Software Engineering intern — Huawei R&D UK

C/C++ · secure element operating system, low latency cryptographic fun

3f0c8b

Software Engineering intern — Certara (UK)

some good old fashioned automation of repetitive tasks in python, the intern classic. 90% latency reduction isn't half bad though.

$ git log --oneline school/

1e88a3

MEng Electronic & Information Engineering — Imperial College London

Systems, signal processing, machine learning, computer architecture, compilers. All in the pre vibe coding era, i.e building a C compiler from scratch sophomore year without Claude pitching in. Feels good.

$ ls -la projects/

drwxr-xr-x2026-08itchy/ NASDAQ ITCH decoder + L3 order book in C++20. 268.7m messages at 2.54m/s, 5.9× faster after some profiling. scratchy is next drwxr-xr-x2026-04the-rusty-vault/ rust/tokio multi-platform crypto gateway. Ingests + normalizes market events over websockets. drwxr-xr-x2025-11ball knowledge/ vote on sports takes, and ascend the ranked leaderboards of public opinion. Know ball? Prove it. Built for iOS e2e with Dart and submitted to the App Store drwxr-xr-x2025-06compliance-tusk/ EPA compliance paperwork, but agentic

$ tree stack/

stack/
├── serving/
│   ├── vllm
│   └── litellm
├── infra/
│   ├── kubernetes
│   ├── docker
│   ├── rabbitmq
│   └── keydb
├── data/
│   ├── elasticsearch
│   ├── redis
│   └── weaviate
├── cloud/
│   ├── aws (bedrock)
│   ├── azure
│   ├── gcp
│   └── circleci / github-actions
├── ai-dev-tools/
│   ├── cursor
│   ├── claude-code
│   └── codex
└── languages/
    ├── python
    ├── c / c++
    └── java (groovy / kotlin)

$ cat now.txt

Reading DDIA, AI Engineering, Musashi Miyamoto and Thomas Paine. Figuring out what vLLM meant in the patch notes. Adjusting at the same speed that AWS fiddles with the Bedrock Converse API. Riding fast motorcycles. Lifting. Running. Rugby once in a blue moon. Bank fishing the White River. Getting Farsi past the point where I can read Rumi without sounding silly.

$ ./contact

email github linkedin

type help — tab completes, ↑ recalls