This is the incredible story of Nikhil Choudhary, an innovator who went from Mechanical Engineering and manufacturing to becoming an AI expert Data Analyst, driven by the rapidly evolving technology landscape. Working at GEP, a leader in sourcing and procurement, he tackled a critical financial problem: global contracts are static, yet geopolitical tariffs (like those imposed by Trump) and logistics costs change daily, causing massive financial losses. Nikhil’s powerful solution is the Predictive Tariff and Contract Generator – AI powered.
This use case employs machine learning algorithms (like XG Boost and decision trees) and AI agents to process realtime tariff feeds, predict cost volatility, and simulate future scenarios. The system then autowrites new pricing clauses, dynamically updating contracts before projects are completed. This ensures justified rates and transparency between parties.
This innovation has delivered significant cost savings, ranging from 10 to 20%, multiplying profit margins for global companies. Nikhil’s journey proves that successful innovators must continuously learn, stay dynamic, and couple strong domain knowledge with AI Readiness Skills.
Share something about yourself
Let me tell you, my journey; it’s been quite a rollercoaster, starting far away from where I am today. Most people assume I’ve always been a tech guy, but actually, I did my engineering in Mechanical. Imagine that! I was a Project Lead in manufacturing, sourcing materials all the way from Delhi to Pune to make parts for coolers. We mechanical folks had absolutely zero idea how computer science guys made applications or started coding.
But the AI world was evolving so fast! I just had to pivot. I realised, chalo, let’s put my hand in the river and see what this is all about. So, I pursued an MBA with a focus on data marketing and simultaneously did a data science course. That’s how I got deep into machine learning algorithms and artificial intelligence tools.
You know, the computer connection wasn't new. Since childhood, I have always been into computers. I remember back in school, around 2012, we had these computers and printers in the library. Once, the lab assistant couldn't fix a printer-a software and hardware issue. But my friends and? We rebooted the software, did some jugaad, and fixed it! The teacher was actually surprised. From that time, we were known to be very good with computer connections.
After my MBA, I joined GEP, which is a big leader in the sourcing and procurement industry. And that’s where the real action started, where we began developing AI tools for automating small parts of our huge business.
Problem Overview
The issue we decided to tackle was huge, especially in procurement. See, contracts are generally static and made in the past. Let’s say a contract was made in January, right? But then, politics gets involved. Donald Trump, for instance, kept imposing new tariffs, changing numbers every month for different countries. The logistics costs and tariffs change literally every day.
Now, the main thought was: If prices are changing today, why should the contract be living in the past? This gap creates chaos. If you miss this change-say you’re a manufacturer in China making a turbine based on imported goods-and the final cost ends up being way higher than what was marked in the contract, the client won't pay you the extra amount. It means a loss for one of the parties, sometimes both. We were solving this specific pain point.
And listen, solving this manually? Impossible. To analyse the spend files, which have lakhs of rows and countless suppliers, you would need around 20 or even 30 analysts. One person feeds the Excel sheet (yes, people still use Excel in supply chain!), another analyses 2030 trade routes, and another looks at 200+ HS codes (Harmonised System codes for materials) and geopolitical triggers. All that work just to report in a month. My tool automates this process, and you get the answer every week.
Motivation
Honestly, the motivation was clear: stopping financial haemorrhaging and staying ahead of the game. Global leaders like Tata Chemicals, Reliance, and Oracle already have their AI tools ready. Other big players don't want to miss out on AI adoption.
When you deal with global procurement, companies buy in massive quantities. If they spend billions or millions, a 15% tariff change can be financially devastating. Even saving just 3 or 4% by diversifying suppliers gives them a huge advantage. When you can show 2% savings, that multiplies directly into their profit percentage.
I knew I had to be dynamic. Like I changed my journey from mechanical engineer to MBA and then data science, companies needed to adapt too. This project, which we call the AIpowered Predictive Tariff and Contract Generator, ensures transparency and justified rates for all transactions.
Solution Overview The AI Use Case
Our solution is about creating AIenabled dynamic contracts. Since every material has an HS code and is shipped via specific trade lanes, we built a smart data pipeline.
Whenever there’s an official announcement on tariffs, say from the US administration, our system immediately gets triggered. It processes the change and alerts both the supplier and the client. We give them an indicator: "Hey, the price in your region is going to change. Because of this tariff, raw materials are going to be impacted by, say, 5%. You need to restart renegotiating your contract for the coming month".
This stops any surprise losses. They renegotiate before the project is completed, deciding whether to go ahead or make a new contract altogether.
Technical Details How AI is Actually Utilised
Okay, now let’s get a little technical, but I’ll keep it simple for you. Our solution flows through four main stages: Data Injection, AI Classification, Predictive and Simulation Engine, and the Dynamic Contract Generator.
- Data Injection: We feed the system all the raw data: HS code customs data, commodity indexes, freight spot rates, and global tariff feeds, along with geopolitical risk signals. We use tools like Restful APIs, Airbite, and Bride data. APIs are the real gamechangers here because they give us realtime analysis you can't do AI on old data, right? We use a mix of free APIs (which are helpful) and obviously, the paid ones, which perform better.
- AI Classification: We use machine learning graphs to map and cluster similar materials based on their HS codes. Then we classify the risks (like tariff hikes or geopolitical issues) by detecting changes in clauses and regulations. We use algorithms like
XG boost and decision trees to make these layers of clusters.
- AI Agents and Simulation Engine: All these complex steps are orchestrated by AI agents. These agents handle specific decisionmaking. The Simulation Engine uses time series forecasting to predict freight indices and tariff trends-like figuring out the 50% tariff US imposed on India, and then looking for substitute suppliers for the market. It predicts landing cost volatility.
- Dynamic Contract Generator: This is the cool part. Based on all the future scenarios the agents predict, the system autowrites new pricing clauses. It uses logic: "If this, then that; else if this, then that." We deployed this using
Lang chain, Python, and Streamlit. We also utilize generic tools like OpenAI and Google News APIs for extracting necessary information.
Challenges & Failures
Every big project has its stumbling blocks. In nontechnical terms, one major challenge was that our tool was often triggered by small, insignificant news articles that weren't actually required. We had to refine that.
Secondly, classifying the types of materials using HS codes across different global regions was a really difficult task. Plus, we had major legal concerns regarding the algorithm clauses and other specific contract aspects.
But here’s the lesson: Accuracy gets better day by day. The more you train the model, the better it performs. The more we make these agents work, the smarter they become over time.
AI Readiness Skills
Let’s understand how good your Collaboration skills are – Name the persons or organisations you collaborated with for making this project successful.
Collaboration is everything; you can’t do this alone. Although there are many GEP colleagues, if I were to name them all, there would be many. I learned the power of teamwork early on.
I did my data science course at the XLR institute. They used to force collaboration by putting us in random groups of five people who had never met before. It took real courage to collaborate and find a single solution. Some of the friends I worked with were Dhaneshwar from Hyderabad, Abdullah (also from Hyderabad), Subhash from Bangalore, and Bhanu. That projectbased learning is vital.
Let us know how good your ‘Critical Thinking & Creativity’ is – Do you think your solution is the best one for the stated problem? If yes, then why? If not, what is the best solution?
As of now, I genuinely think it is the best solution we can provide for managing dynamic contracts affected by tariffs. It integrates realtime risk simulation and contract generation-that's pretty cutting edge.
That said, there is always scope for improvement. That’s the beauty of this field; it never stays still.
Let’s test your Online Learning skills – What selflearning did your team do to understand the problem better, or design the solution, or for any other aspect of the project? Also, please specify the sources (if possible) from which you learnt.
Selflearning is now a crucial skill. For problemsolving, we started using AI tools like Perplexity, GPT, and Claude extensively.
Here’s a trick: You use prompt engineering to frame your exact problem in the LLM. Then, you ask the application, “How did you do this? Give me the technical terms and the workflow,”. It opens up your eyes and gives you the food for thought you need. It's like talking to AI is the first step.
But before all that, you must have strong basics. My advice is to clear your basic concepts in math calculus, trigonometry, and complex numbers till 11th and 12th standard. Mastering these opens up your brain analytically. You need trigonometry for things like making regression models and calculating slopes. If you don’t know why you are studying these subjects, ask an AI tool: "Where is the actual application of this numerical in the real world?". It will show you how sin theta and cos theta are used in oil rigs or manufacturing plants, making the learning meaningful.
Let’s test your Online Reputation skills – If you have made any effort to popularise your solution on the Internet, please share the relevant links.
Since the Predictive Tariff and Contract Generator is our company’s intellectual property (IP), I cannot share any specific links or demo videos for this particular project.
However, I have used my own AI use cases to popularize the concept of automation. For example, I once loaded a huge monthly spend file (lakhs of rows) into my Python notebook. I fed a sample file to Perplexity and asked it: "I want to analyze this spend data and show it to senior leadership in one hour. Give me the entire Python code for the analysis and save it as a PPT with visual charts". Within minutes, I had the complete PPT ready. That’s the kind of impactful automation that resonates online.
Impact so far & vision for the future
The impact has been significant, yaar. We’ve managed to put contracts in place that were previously missing, and we’ve generated cost savings ranging from 10 to 20%. When a global company is spending in millions and billions, a 10% saving directly adds to their revenue. Even saving 2% multiplies the profit percentage.
For the future, I see this accelerating the job role changes. Earlier, if a company needed 10 data analysts, now maybe just two can handle everything, provided they have AI upskilling and reskilling. Domain knowledge is crucial, but coupling it with the ability to build agentic AI pipelines is what keeps you in the race.
Your advice to fellow innovators
My advice is simple, but powerful: Keep your mind open.
Start your day by reading the newspaper. No AI tool can replace the habit of reading the news. You need to understand what is going on globally, in the market, first. Once you know the problems, you can identify solutions and start building.
For staying updated, I highly recommend checking out TechCrunch. They publish all the latest updates on AI, including announcements from Sam Altman and GPT.
Mentor, Support & Inspiration
I am highly inspired by the senior leadership at my company, GEP. They are truly open to advice and give amazing freedom to your thoughts. Even if you are the most junior person in the meeting, your input is considered. That kind of environment motivates you to think bigger and innovate.
Nikhil Choudhary demonstrated another highly practical AI use case focusing on AIPowered Procurement Analytics. This solution addressed the severe time constraint faced by procurement managers who typically spend days analyzing large spend datasets, often consisting of lakhs of rows and hundreds of columns, using slow, traditional methods like Excel.
His approach utilizes intelligent automation to provide Cost Savings Insights in seconds. Nikhil used prompt engineering, feeding the AI tool (Perplexity) sample column headers and asking it to generate endtoend Python code for visual analysis and to save the results directly in a PPT format. In moments, the code produced detailed charts on procurement spend distribution, price anomalies, and, critically, an Executive summary of procurement cost savings opportunities.
This use case is highly insightful for others because it clearly demonstrates the power of AI to convert days of manual labor into mere seconds of output. The economic insight is profound: as Nikhil explains, saving just 1% to 2% of cost goes directly into profit, potentially increasing overall profit dramatically when dealing with millions and billions of dollars. Furthermore, the method itself using AI to generate complex technical workflow instructions serves as a strong example of effective Online Learning and leveraging AI tools to quickly design solutions and accelerate execution.