SRI JAVALI KOTHA
Computer Science undergraduate · AI/ML, Generative AI, Computer Vision
Hyderabad, Telangana, India
@sri_javali_kotha
0 followers
B.E. Computer Science & Engineering at Keshav Memorial Engineering College · Graduating 2027
About
AI/ML-focused Computer Science & Engineering undergraduate with hands-on experience in machine learning, deep learning, computer vision, Generative AI, LLMs, RAG, and agentic systems. Has research experience at IIIT Hyderabad and product engineering experience building AI-enabled applications and production-oriented APIs.
Experience
-
Research InterniHub-Data, IIIT Hyderabad · Hyderabad, Telangana, IndiaMay 2025 – Aug 2025
During my Research Internship at IIIT Hyderabad, I worked on a YOLO-based computer vision system for object detection in dense crowd-safety scenarios. I was involved in data preprocessing, model experimentation, training, evaluation, and optimization, with a focus on improving detection robustness under varying crowd densities and occlusion. I also analyzed false positives, accuracy, latency, and reliability to understand the trade-offs involved in real-time deployment.
-
Research InternIIIT Hyderabad
Conducted computer-vision research on real-time crowd-safety analysis using YOLO-based object detection, focusing on robustness under crowd-density variation and occlusion. Prepared datasets, ran experiments, evaluated model behavior, and analyzed accuracy, latency, false positives, and reliability under real-time deployment constraints.
-
Full Stack (AI/ML) Developer InternPM Accelerator
Developed production-facing AI-enabled product workflows across buyer onboarding, buyer profiles, acquisition preferences, seller profiles, and business workflows in MatchBook. Integrated frontend workflows with backend APIs, authentication, Supabase persistence, typed data contracts, validation, state management, and error handling. Collaborated through Git branches, pull requests, reviews, merges, and iterative requirements while debugging TypeScript and frontend/backend integration issues.
-
Independent ResearchFreelance
Designing a multi-stage, model-agnostic guardrail framework addressing prompt injection, jailbreaks, retrieval abuse, sensitive-information disclosure, and data exfiltration. Engineering SAFE/HIGH_RISK trust classification with intent analysis, retrieval-risk assessment, contextual safety, confidence scoring, and interpretable explanations; exploring LoRA fine-tuning and retrieval governance.
Education
-
2027
B.E.Keshav Memorial Engineering CollegeComputer Science · 2027
Skills
Tools / apps / platforms
Projects
-
Swasthya — Healthcare Information PlatformHealthcare Information Architecture
Built a structured healthcare information platform emphasizing accessible organization, clear navigation, and user-oriented presentation of health resources.
-
ICU Mortality Prediction — Clinical Machine LearningMIMIC-III, MIMIC-IV, eICU, CatBoost, LightGBM, AdaBoost
Developed leakage-aware clinical ML workflows for ICU mortality prediction across heterogeneous critical-care datasets, harmonizing clinically related variables, units, missingness, and temporal information. Compared ensemble and neural approaches using reproducible train/validation/test splits, preprocessing pipelines, temporal features, and model evaluation focused on clinically meaningful prediction.
-
JoyVerse — AI-Powered Learning System for Dyslexic ChildrenReact, FastAPI, WebSockets, Vision Transformers, MediaPipe
Built a real-time AI learning platform combining computer vision, Vision Transformers, MediaPipe Face Mesh, WebSockets, and learner dashboards for adaptive interaction and progress reporting. Integrated visual signals and facial-expression analysis into an application workflow designed for personalized and accessible learning experiences.
-
Leaf Disease Detection & Advisory System (JARNS)Python, Vision Transformers, FastAPI, RAG, ChromaDB, OCR
Built an end-to-end AI application combining Vision Transformer disease detection, FastAPI inference, RAG, ChromaDB semantic retrieval, OCR, weather data, and government RSS feeds. Implemented image-data cleaning, normalization, augmentation, validation, embedding-based retrieval, multilingual advisory generation, and automated knowledge-base synchronization.
-
Atlas AI — Proactive Financial AI AssistantPython, FastAPI, PostgreSQL, LLM Tool Calling, Telegram, Pytest
Built a persistent agentic financial assistant that understands goals, gathers context, selects tools, reasons over structured data, performs actions, persists state, and schedules follow-up work. Integrated Razorpay merchant payment data to analyze payment failures, estimate revenue at risk, identify recovery opportunities, and generate proactive financial insights.
Extracurricular activities
-
Student Coordinator, Computer Society of India (CSI)
Student leadership role in CSI.