KRISHNA RAO KAKKIRALA
GenAI | Agentic AI
Hyderabad
Summary
16+ years in AI/ML development, with 8+ years leading high-impact projects and cross-functional teams. Expertise: Developing scalable AI solutions in Computer Vision, NLP, IoT data processing, Machine Learning, Deep Learning, and Generative AI (LLMs & AI Agents) while optimizing models for efficiency and performance. AI Architecture & Deployment: Designing cloud-native AI solutions on Azure, AWS, and GCP, integrating LLM frameworks (LangChain, LlamaIndex) for seamless deployment. Leadership & Innovation: Driving AI research, patents, and publications while mentoring teams and ensuring MLOps best practices for continuous model improvement. Business Impact: Enhancing strategy through Predictive Modeling, Data Science, and AI-driven analytics, fostering strong stakeholder collaboration for transformative AI solutions.
Overview
17
17
years of professional experience
2008
2008
years of post-secondary education
2
2
Languages
Work History
Associate Research Scientist (Sr. Lead and Developer)
Toshiba
12.2018 - 02.2025
- Developing a Financial AI Agent to retrieve relevant insights from extensive financial documents based on user queries. Overcoming challenges like extracting structured data from tables for accurate interpretation and contextual relevance.
- Developing a Video Analytics Agent for the advertisement industry using CLIP ViT model, Claude, PyTesseract, Deepgram, and Transline.
- Developing a Database AI Agent that processes text-based queries to interact with databases like SQLite, MySQL, and PostgreSQL, retrieving relevant data and presenting output in an easy-to-understand visual format.
- Developing an Expense AI Agent leveraging OCRs and LLMs to extract desired information from scanned documents.
- Led the development of a Chatbot using Deep Learning and Generative AI techniques, enhancing user interaction and support.
- Extracted text from images using OCR algorithms (Tesseract, EasyOCR, Llamaparse) and applied Natural Language Processing (NLP) methods to populate fields in forms.
- Developed Text Extraction Methodologies to extract relevant text from specification documents based on specific queries, increasing data accessibility.
- Extracted PDF Data including text, tables, and figures for use in LLMs and Retrieval-Augmented Generation (RAG) systems, enhancing query responses.
- Led the successful AI Integration of advanced language models to develop cutting-edge AI solutions, utilizing frameworks such as LlamaIndex.
- Successfully led a team of up to 14 members, actively contributing to various data science and machine learning projects.
- Generated Synthetic Defective Integrated Circuit (IC) images using CycleGAN and StyleGAN, facilitating model training and anomaly detection. Also experimented with diffusion models such as ControlNets.
- Developed a project to generate VLSI chip thermal images with different faults using Flux model and LoRA fine-tuning.
- Designed and developed methods for segmenting grain regions and determining sizes in steel images using Mask R-CNN and image processing techniques.
- Introduced incremental learning, YOLO, and machine learning techniques for segmenting and classifying crimp images.
- Built robust classification algorithms for datasets with mislabeled and highly imbalanced data using CNNs and Vision Transformers (ViTs).
- Identified anomaly regions in images using weakly supervised and unsupervised learning techniques, including SPADE+, GCPF, and IFNET.
- Implemented Medical Image Processing algorithms for tasks such as measuring patient dimensions over CT scan couches and deploying these solutions.
- Developed Anomaly Detection Algorithms based on Acoustic Emission (AE) signals using Wavelets and Statistical features, improving defect detection accuracy.
- Evaluated and identified effective tuning parameters using HyperOpt, optimizing model performance.
- Developed deep learning model compaction using AiP-based methods and OpenVino, achieving a 50% reduction in model size.
- Specialized in Zero-shot learning, unsupervised learning, weakly supervised learning, supervised learning, and incremental learning.
- Deployed data-driven solutions using advanced Machine Learning, Deep Learning, and AI technologies.
- Formulated comprehensive reports and dashboards to present data insights and project outcomes.
Researcher
TCS Innovation Labs, Tata Consultancy Services
07.2008 - 11.2018
- Designed & developed Robust watermarking algorithms to protect the copyrights of videos and speech books.
- These algorithms are developed in C++, OpenCV.
- Which are invariant to various kinds of distortions.
- Robust Face Recognition Algorithm to recognize cab drivers by utilizing Computer Vision and Machine Learning Algorithms.
- The algorithm is robust against occlusions, partial faces.
- Cross-Model (Thermal to Visible) Face Recognition Algorithm using Deep Learning.
- Visual Target Tracking Algorithms using Deep Learning for Drone surveillance.
- Biometric Template Protection Algorithm using Deep Learning.
- Achieved automatic masking of fields for scanned documents to protect privacy of a customer using computer vision and machine learning algorithms.
- Take care of GDPR guidelines.
- Sequence prediction of Visual-IQ Questions via Deep Learning.
- Automatic answering a natural language question against a video using Deep Learning for surveillance video applications.
- Image watermarking to automatically extract expiry information when images are posted on websites.
- Invisible watermark.
- Robust against compression.
- Every day, the application automatically tracks all the images posted on websites and checks expiry data to know whether it meets terms or not.
- Discard images from publish if image meets its contract end date.
Education
M.E. - Communication Systems
PSG Technology
Coimbatore, Tamil NaduB.Tech. - Electronics & Communication Engineering
Jawaharlal Nehru Technological University (AGCET)
Tadepalligudem, A.P.Skills
Collaborator
Lead
Coach
Developer
Communicator
Innovator
Analyzer
Planner
C
C
Python
TensorFlow
Pytorch
Flask
FastAPI
TensorRT
CuDNN
Pyspark
MySQL
NoSQL
GIT
AWS
MicrosoftAzure
OpenCV
MLFlow
MLOps
ML
DL
LLMs
Llama Index
Langchain
LLM Finetuning
Accomplishments
- Merit of granted 18 Patents in (TCS, Toshiba) (US, Japan, India)
- Conferred with: Best Paper Award from TCS Tactics in 2011
- Best Innovation Award from TCS Innovista in 2017
- Successfully selected two innovative ideas for the Toshiba Ideathon 2024
Personal Information
Date of Birth: 07/12/83
Timeline
Associate Research Scientist (Sr. Lead and Developer)
Toshiba
12.2018 - 02.2025
Researcher
TCS Innovation Labs, Tata Consultancy Services
07.2008 - 11.2018
B.Tech. - Electronics & Communication Engineering
Jawaharlal Nehru Technological University (AGCET)
M.E. - Communication Systems
PSG Technology