Aditya Manoj Krishna

Evaluated and trained computer vision models for a biped robot for the Defence Research and Development Organisation (DRDO), including segmentation, detection, tracking, and NIR-based temperature prediction.
Designed and built the Robot prototype for in-lab testing of the vision system.
Ran tests of the models in different scenarios to understand their performance and limitations in real world scenarios.

Designed and optimized a chatbot system initially with a 5-minute query runtime for a document corpus of approximately 100,000, reducing it by 99.8% to 6 seconds.

Developed both versions independently using Langchain, Chroma (as the vector database), Azure OpenAI, and Python.

Implemented text-splitting algorithms and serialization techniques to embed, store, and retrieve relevant text chunks from the corpus.

Contributed to the backend team responsible for developing software for a watch designed to process sensor data. This included monitoring ex-inmates on probation for vital signs such as heartbeat, oxygen level, location, and more.

The software increased data processing speeds and reliability by 67%, allowing for reduced probationary officer visits.