I'm an MS Computer Science student at USC. Before that, I spent a year on the AIML team at Pennant Technologies, building AI agents that talk to customers over chat and phone, and the data pipelines that feed machine learning models in production.
Now: building an end-to-end AI personal assistant that automates everyday tasks, like JARVIS.
I'm most excited about agentic AI: systems that don't just answer a question but carry a person through a whole task. At Pennant, I worked on two of them, a chatbot agent and a voice agent that answers phone calls, both guiding customers through a loan application from start to finish.
I also like the unglamorous half of machine learning: getting clean data to the model on time, and getting predictions back to the people who need them. That meant ETL pipelines on Apache Spark and Livy feeding AutoML training and prediction serving for enterprise clients.
I'm looking for roles in software engineering, machine learning, data engineering, and AI, wherever systems have to be reliable and wired into real work. Outside of work, I've co-authored research on multilingual lip-sync video, built a tool that measures the carbon cost of LLM inference, shipped an Android app, and helped run a 500-person inter-college hackathon.
Applying for a loan is a long, multi-step process: personal details, employment and income, documents, and questions that change depending on earlier answers. Forms like this are where applicants get confused and give up. The goal was to replace the form with a conversation, so a customer could simply talk their way through the application.
I built agents for a customer-facing chatbot that guides applicants through the loan application step by step. I also worked on a voice agent that answers incoming phone calls and helps callers with the same process, so customers who would rather talk than type get the same help.
Both agents do more than talk. They call internal lending APIs to check a customer's eligibility and the status of their application, so answers come from the company's real systems rather than guesses.
Swipe to see the full diagram.
Building these agents is why I'm so interested in agentic AI. A chatbot that answers questions is useful, but an agent that carries someone through a whole task, across channels, is a different kind of software. That's the idea behind the personal assistant I'm building now.
Every API call to a generative model burns energy somewhere. EcoTrace tracks inference in real time and estimates energy use and emissions from computational load, server hardware, and the carbon intensity of the region it runs in. Developers get token-level emission figures and aggregate reports, so they can see which prompts and models cost the most and cut waste.
An AI-driven app for tracking meals and managing meal tokens, built on AWS services. Our team placed second.
A pipeline that takes a video in one language and produces it in any of 100+ others. It translates the speech, generates the new audio, and re-syncs the speaker's lips so the result looks naturally spoken. The output reached about 90% accuracy and became the basis for our IEEE paper, MultiLingualSync.
A live website for a national-level go-kart endurance championship hosted at our college, giving participants and spectators real-time information throughout the event. The college asked us to take it on at the last minute, after the original team stepped away. We designed and built it overnight and had it live before registrations opened.
Built from my own commute: students were marking each other present on the college bus. The app verifies attendance with a QR scan plus geofencing and GPS checks, so a scan only counts if you're actually on the bus. Logs are tamper-resistant and sync with WebPros India's backend APIs.
MultiLingualSync: A novel method for generating lip-synced videos in multiple languages
Nanditha G., Datla K. V., Kevin G., Nikitha R., Pallavi L., Babu C. M. 2023 3rd Asian Conference on Innovation in Technology (ASIANCON), IEEE.
Data analytics on opportunities for women in the field of technology
Pallavi L., Kosuru S. S., Datla K. V., Debbata K., Dulam A. G., Gangavarapu R. C. 2023 5th International Congress on Human-Computer Interaction, Optimization and Robotic Applications (HORA), IEEE.
Our team won second place and a ₹75,000 prize for an AI-driven meal tracking and token management app built on AWS.
Cloud Foundations, AWS Academy, 2022
Android Development with Kotlin, Google for Developers, 2022
Have a question about my experience? Ask this agent. It answers from my resume and this site, and it will say so when it doesn't know.
I'm open to internships and full-time roles in software engineering, machine learning, data engineering, and AI.
kaushik.datla@gmail.com