Welcome to the Github profile of Vamsi Krishna Kocherla. I am a Senior Computer Vision Scientist with a strong enthusiasm for AI and Robotics, currently based in the United States.
With a Bachelor's degree in Electronics and Communication Engineering and a passion for coding, I have focused my career on artificial intelligence. I am currently expanding my expertise by pursuing a Master of Science in Robotics at the University of Minnesota - Twin Cities. My experience includes developing complex detection and classification models and designing microservice-based infrastructures for deployment, reflecting my skills in applying theoretical knowledge to practical challenges.
Side Quests: Learning hardware architectures, trying to break and fix stuff, always in search of things I do not understand, to one day make sense of them and connect all the dots.
End Goal: Invent something so profound and satifying, that I finally feel like taking rest.
Behold! The meticulously curated (and oh-so-humble) display of skills I've collected like digital Pokémon.
| 💪 Core Skills | 🚀 Programming Languages | 🔮 ML & DL Libraries |
|---|---|---|
| Deep Learning | Python | PyTorch |
| Machine Learning | C++ | TensorFlow |
| Computer Vision | Golang | OpenCV |
| Robotics | CUDA | Scikit-learn |
| GenAI | NodeJS | Transformers |
| Neural Networks | CNNs, GANs | |
| Image Processing | TensorRT, ONNX | |
| NLP | ||
| SLAM | ||
| 3D Reconstruction | ||
| Stereo Vision |
| ⚙️ DevOps, Cloud & Deployment | ☁️ Cloud Platforms | 💽 Databases & Tools |
|---|---|---|
| Kubernetes | Google Cloud Platform | MongoDB |
| KNative | Amazon Web Services | MySQL |
| KServe | Azure Cloud | PostgreSQL |
| Triton Inference Server | Redis | |
| Docker | ClickHouse | |
| Kubeflow | RabbitMQ | |
| Flask | Apache Kafka | |
| Fast API | Prometheus, Grafana | |
| REST APIs, gRPC, WebSockets | Git, DVC | |
| Gitlab CI/CD | Ubuntu, Linux File System | |
| Linux | Matplotlib |
| 🤖 Robotics & Hardware |
|---|
| ROS/ROS2 |
| Isaac Sim (Omniverse) |
| Gazebo |
| OpenAI Gym |
| Ardupilot, PX4 |
| MQTT |
| Jetson Orin Nano, AGX |
| Raspberry Pi |
| Arduino |
| Sensor Fusion, Integration |
- Building microservice architectures like they're Lego castles.
- Taming GPU workloads in Kubernetes clusters (it's mostly screaming internally).
- Deploying ML models to tiny edge devices, because who needs resources?
- Making inference faster and cheaper, often involving arcane rituals with FP16/INT8 precision on fancy GPUs.
- Turning research papers into functional code (sometimes it even works!).
- Automating the boring stuff with CI/CD pipelines.
- Mastering the art of reading inscrutable AI research papers.
- Wrangling HPC Systems (High-Performance Computing, not Hugely Problematic Contraptions... usually).
- Making things blink with Arduino, Raspberry Pi, and other embedded fun.
- Occasionally pointing a camera at things and pretending I'm artistic (Adobe Lightroom helps).
- Upjao Agrotech (Sr. Computer Vision Scientist, Core Team): Battled agricultural data, improved model accuracy by up to 20%, built a serverless inference beast, shrunk models for edge devices, and made things go brrr (faster inference). (Jan 2019 - Jul 2024)
- Tata Consultancy Services (AI/ML Developer): Survived the elite TCS Rapid Labs Cohort, whipped up NLP POCs like a short-order cook, and devised algorithms to deduplicate messy data. (Jun 2021 - Apr 2023)
Need the boring professional version? Stalk me on LinkedIn.
- University of Minnesota - Twin Cities: Master of Science in Robotics (Expected Aug 2026) - Because one degree clearly wasn't enough punishment. Coursework includes Computer Vision, Intelligent Robotic Systems, Image Processing, NLP, Deep Learning.
- GITAM University, Hyderabad: Bachelor of Technology in Electronics & Communication Engineering (Jun 2017 - Jun 2021) - Where it all began. Relevant suffering included Signals & Systems, DSP, Probability Theory.
- Autonomous Drone Navigation with Stereo SLAM: Made a drone see in 3D with custom stereo vision and navigate without GPS using things like ORB-SLAM3/RTAB-Map. Involved ROS 2, Jetson Orin Nano, Pixhawk, and a healthy dose of hoping it wouldn't crash. Tested in simulation first, because real-world crashes are expensive.
- Image-Based Visual Servoing (IBVS) for Robotic Arm Control: Taught a robotic arm to grab things using visual feedback. Built a stereo camera system, generated depth maps (shoutout to the "Stereo Anything" paper and YOLOv8), and optimized it with TensorRT for speed.
- DIY Drone (Quadcopter): Because buying one is too easy. This beast could supposedly fly higher than a seven-story building, had a 2.8km range, 20 min flight time, and could lift 6kg. Built with Pixhawk, ESCs, motors, and probably some duct tape.
- Image Super Sampling (SRGAN): Turned a 720p movie into 4K using dark magic (SRGAN). Involved fancy memory management to avoid GPU meltdowns.
- Real-Time Noise Cancellation: Tried to silence the chaos using Deep Learning (Wave-UNet). Fed it clean speech and horrible noise to teach it the difference. Achieved an MSE loss of ~0.2, which sounds vaguely impressive.
Age Prediction Server(Let's pretend this wasn't here, the resume doesn't mention it, must have been a fever dream).
Want to see the code behind the curtain? Dare to explore my GitHub.
- Yes, there's a paper on Noise Cancellation.
- And patents! Stuff about assessing agricultural products, encoding data with markers, acquiring images uniformly, classifying with less data (filed), and even identifying cattle by their faces (filed). Don't ask.
So, feel free to poke around. If you have questions, collaboration ideas, or just want to debate the merits of PyTorch vs TensorFlow, my virtual door is slightly ajar. Send an email to koche156@umn.edu 👋👋