Agentry
Book a scope call
About

Founder-led. Field-tested.

Agentry exists because most AI projects fail in the gap between a demo and a deployment. We’ve spent years on the deployment side of that gap — with clinicians, analysts and operators — and now we do it for startups and mid-size teams that can’t afford a science project.

Jawad Amir, founder of Agentry
JA
Jawad AmirFounder · Senior AI Engineer · B.Sc. CSE, RUET
The person behind it

Seven years shipping software. Three building agents people rely on daily.

Jawad is a customer-facing engineer with seven years of production software behind him, the last three focused on LLM and multi-agent systems. He works the whole path: sit with the customer, learn the domain and the data, prototype fast against the real workflow, then harden it into a governed, observable deployment — and stay on after go-live.

Most recently he was forward-deployed on a multi-agent analytics platform used daily by non-technical business teams; before that, embedded with practising clinicians at IQVIA, where the system he built cut research time by roughly 40%; and before that, delivering 5+ end-to-end projects for European enterprises at Selise. Comfortable across Python, TypeScript, .NET, AWS and Azure — and equally comfortable in the room while requirements are still ambiguous.

7+years production software
3years LLM & multi-agent
4companies, 3 countries, all remote
LinkedIn GitHub Ask the CV assistant
How we work

Four rules we don’t bend.

01

Built with you, not delivered over the wall

We shadow the real work before we build, and your team uses the prototype while it’s still cheap to change. The clinicians, analysts and operators we’ve built for are the reason the systems got adopted.

02

Governed by infrastructure, not prompts

“Please don’t do X” in a system prompt is not a security control. Permissions live at the gateway, risky actions wait for a human, and every run leaves a trace.

03

Evaluated before customers see it

Hundreds of graded scenarios run on every change. If a prompt or model update regresses behaviour, CI goes red — not your inbox.

04

You own everything

Your cloud, your repo, your accounts, your data region. We hand over infrastructure-as-code, the eval suite and runbooks. Keep us on or don’t — it works either way.

Track record

Where the patterns come from.

Everything we build for clients was first shipped in production somewhere real. This is where.

  • Feb 2026 — now

    Anlytic · Dubai (remote)

    Forward Deployed / Senior AI Engineer — multi-agent analytics platform

    Built the agent layer on Anthropic Claude and the Vercel AI SDK: a Haiku intent classifier routing to specialist agents for chart generation, dashboard editing, data management and analytics Q&A. Cut LLM spend 60–70% through context engineering, built a 350+ scenario eval release gate, migrated the tool layer onto Amazon Bedrock AgentCore Gateway as MCP tools, hardened authorisation with Cedar policies, and added human-in-the-loop approval for data-sensitive actions.

  • Sep 2024 — Feb 2026

    IQVIA · Fortune 500 (remote)

    Senior Software Engineer — AI clinical decision support & enterprise health-data platform

    Built a RAG pipeline over millions of medical papers and clinical studies (Azure OpenAI, Pinecone) shaped around how clinicians actually search — research time dropped ~40%. Shipped clinical decision support with evidence-based recommendations and drug-interaction warnings inside the daily workflow, ML anomaly detection that improved trial-data accuracy 35%, and .NET 8 microservices serving millions of patient records on AWS EKS.

  • 2024 — 2025

    Autonomix · Denmark (remote)

    AI Software Engineer (part-time) — Donna, a multimodal AI executive assistant

    Core engineer on an assistant automating email, Slack, WhatsApp and calendar workflows on the Claude Agent SDK — 76 tools across 9 business domains. Designed a progressive memory system that keeps per-request memory cost flat, agentic RAG over cross-platform communications (pgvector hybrid search, Neo4j graph retrieval), and relationship intelligence that adapts tone per contact.

  • 2019 — 2024

    Selise · Zurich (remote)

    Software Engineer — client-embedded delivery for European enterprises

    Delivered 5+ projects end to end, on-site with client teams. Built a configurable business-metrics platform from scratch that generated roughly $2M in revenue, a multi-level-query Hub System on microservices that lifted annual revenue 10%, and tuned RabbitMQ to halve latency and raise throughput 12×.

  • 2015 — 2019

    Rajshahi University of Engineering & Technology

    B.Sc. in Computer Science & Engineering

    3rd place, Intra-RUET Programming Contest 2019 · Top 20, IUBAT National Collegiate Programming Contest 2019 · 280+ LeetCode problems · Codeforces 1279.

Work with us

Bring the workflow. We’ll bring the patterns.

Thirty minutes to map one process together. If AI isn’t the right fix, we’ll say so.

Book a discovery call or email xawadamir0@gmail.com