Orchestrate

Mrinal DesaiData Scientist

Winner — HackerRank Orchestrate Global Hackathon, Rank #1 of 1,983. I finished #1 in the August 2026 edition with a final score of 84.2/100. HackerRank reported 22,000 builders signed up globally, with 1,983 completing the AI interview. I built a personalized notification router that decides whether an incoming WhatsApp message should notify, go to digest, or be muted. The key idea was simple: let the model describe, let deterministic code decide. Thank you to HackerRank for a global challenge that tested not only what we could build with AI, but whether we understood and could defend what we built.

Muhammad Rifqi HaikalMedical Doctor

#3 out of 22,000+ registered developers at HackerRank Orchestrate. 24 hours to design, build, and ship an AI agent. 2,475 people shipped something. 1,983 made it through the AI interview. Final score 83.0, with the top three separated by 1.2 points. The interesting part of these 24-hour formats is not the coding. It is deciding fast what not to build.

Antoine MouawadAI Automation & Systems

During one intensive day, I took part in HackerRank's Orchestrate – August Edition, an AI engineering competition focused on building production-ready AI agents. The challenge went far beyond prompting an LLM. It required designing an agent capable of reliable decision-making while handling prompt injection, multimodal inputs, historical context, deterministic policies, and structured outputs. It reinforced something I increasingly believe: building reliable AI systems is as much about software engineering, evaluation, and architecture as it is about the models themselves.

Jeffrey ChuEngineering Research Assistant, New York Institute of Technology

A few weeks ago, I had the opportunity to participate in HackerRank Orchestrate, a global AI-based hackathon oriented around AI development, prompt engineering, and automation. Throughout the hackathon, I designed an AI-powered SMS system to review all of the information associated with a user and a message, then immediately notify, defer, or reject messages depending on their importance to the user. This result placed me in the top 10% of all global submissions. I learned a lot about agentic development, especially how powerful AI tools can be when given a good plan to work with.

Eric PhillipsComputer Science Graduate, Penn State University

It was a lot of fun competing in HackerRank's August Orchestrate competition, where I was tasked with building a personalized message notification router for WhatsApp. I made a video describing how it works, some bugs I experienced, and the lessons learned. It was my first time competing and submitting; I got 79th of 1,983 global submissions, and 10th in the US. My thanks go out to HackerRank for putting on a great hackathon, and congrats to my competitors!

Srikanth Reddy NandireddyAI Engineer

temperature=0.0 did not make my pipeline reproducible. Same model, same prompts, same decoding parameter — 43 of 140 readings came back different. Determinism is a property your deterministic layers have. It is not a property of your system end to end, and the temperature parameter will not give it to you. Across two editions of HackerRank Orchestrate — a 24-hour solo build where an AI judge reads your repository, then interviews you about it for 30 minutes — everything I got wrong traces back to measuring something adjacent to what was actually graded.

Saai SyvendraComputer Science & Engineering

I entered HackerRank Orchestrate, a global hackathon, mostly out of curiosity. I wanted to see how HackerRank would run a hackathon that openly tells you to use AI. I ended up winning it and learning a lot more than I expected. The task: build a terminal support agent for HackerRank, Claude, and Visa in 24 hours, then defend it to an AI judge. 12,885 people registered, from around 50 countries.

Bhavya KhatriAI Engineer building agentic systems @ Mindtickle · HackerRank Orchestrate Top 2

I came 2nd. HackerRank Orchestrate wasn’t a typing-speed contest. It graded four things end-to-end: the agent you built, the tickets it actually handled, the way you directed your coding tools, and a 30-minute live defence in front of an AI judge. That last one mattered. It is one thing to ship an agent; it is another to explain — under questioning — why your routing, your memory, and your fall-back paths are the right ones. Agentic engineering is becoming a real, measurable skill. "Used Copilot" is not the bar anymore. Designing the loop is. Top 2 of 12,885 across 48 countries is a moment to mark. Then back to work — the standard moves fast.

Mrinal DesaiData Scientist

Winner — HackerRank Orchestrate Global Hackathon, Rank #1 of 1,983. I finished #1 in the August 2026 edition with a final score of 84.2/100. HackerRank reported 22,000 builders signed up globally, with 1,983 completing the AI interview. I built a personalized notification router that decides whether an incoming WhatsApp message should notify, go to digest, or be muted. The key idea was simple: let the model describe, let deterministic code decide. Thank you to HackerRank for a global challenge that tested not only what we could build with AI, but whether we understood and could defend what we built.

Muhammad Rifqi HaikalMedical Doctor

#3 out of 22,000+ registered developers at HackerRank Orchestrate. 24 hours to design, build, and ship an AI agent. 2,475 people shipped something. 1,983 made it through the AI interview. Final score 83.0, with the top three separated by 1.2 points. The interesting part of these 24-hour formats is not the coding. It is deciding fast what not to build.

Antoine MouawadAI Automation & Systems

During one intensive day, I took part in HackerRank's Orchestrate – August Edition, an AI engineering competition focused on building production-ready AI agents. The challenge went far beyond prompting an LLM. It required designing an agent capable of reliable decision-making while handling prompt injection, multimodal inputs, historical context, deterministic policies, and structured outputs. It reinforced something I increasingly believe: building reliable AI systems is as much about software engineering, evaluation, and architecture as it is about the models themselves.