Hi! I'm Balaji Kartheek 👋
I'm an Conversational AI Engineer at Avaamo, building enterprise RAG systems, LLM evaluation frameworks, and multilingual AI platforms. I engineer production-grade pipelines with RAGAS, LangChain, Selenium automation, and experiment tracking via MLflow and Opik.
Outside work, I built JanVaani — a multilingual AI assistant that helps citizens discover government welfare schemes using RAG, semantic search, and vector embeddings — deployed live on Vercel with a FastAPI backend and Supabase/PostgreSQL stack.
About Me
ML Engineer experienced in developing enterprise Conversational AI platforms, Retrieval-Augmented Generation (RAG) systems, and AI evaluation frameworks. Skilled in building LLM evaluation pipelines, multilingual AI capabilities, and scalable automation solutions while improving model quality, scalability, and user experience.
What I Work On
- LLM Evaluation & RAG — End-to-end evaluation frameworks, RAGAS metrics, batch processing, and reporting pipelines for enterprise knowledge bases.
- Enterprise Conversational AI — Multilingual support, document ingestion, PII masking, citation management, and intelligent fallback responses.
- Test Automation at Scale — Python + Selenium + Behave frameworks across UI, REST API, IVR, and multi-channel conversational flows.
- Applied AI Products — Full-stack RAG applications from backend APIs to vector search and live deployment.
Work Experience
Conversational AI Engineer @Avaamo
Jan 2024 – Present · Bengaluru, Karnataka
- Engineered the LLaMB Regression Evaluation Framework for enterprise RAG applications, automating end-to-end LLM evaluation using RAGAS, batch processing, and reporting pipelines — reducing evaluation costs by ~50%.
- Enhanced enterprise Conversational AI with RAGAS-based evaluation (Faithfulness, Answer Correctness, Statement Simplification), multilingual support, document ingestion (CSV, Excel, PPTX), PII masking, secure citation lifecycle management, and intelligent fallback responses — improving platform reliability, security, and user experience.
- Architected a scalable automation framework using Python, Selenium, and Behave, automating 800+ UI, REST API, IVR, and multi-channel conversational AI test cases — significantly improving regression coverage and release validation.
- Accelerated sanity and regression execution by 10× through Python multithreading, reusable automation pipelines, configurable cross-channel validation, and integration with MLflow and Opik for AI evaluation and experiment tracking.
Computer Vision Intern @ Intel Corporation
Jun 2023 – Jul 2023 · Remote · Internship
- Built a real-time social distance monitoring system using Intel OpenVINO, achieving 3× higher inference throughput on Intel hardware.
- Optimized computer vision inference pipelines with OpenVINO, improving FPS by 40% over the baseline TensorFlow implementation through hardware-aware optimization.
- Integrated object detection and multi-object tracking to enable real-time people detection and distance estimation for efficient edge AI deployment.
Machine Learning Intern @ Feynn Labs
Jan 2023 – Mar 2023 · Remote · Internship
- Built and trained ML models using Keras for applied business use cases, from problem scoping through baseline deployment.
- Performed data preprocessing, feature engineering, and model evaluation to improve prediction quality on structured client datasets.
- Compared model variants using standard metrics (accuracy, precision, recall, F1) and iterated on hyperparameters to reach stable performance.
AI Member @ IE Mechatronics Students' Chapter, Manipal
Jun 2021 – Jan 2022 · Karnataka, India
- Contributed to AI initiatives using Python and MATLAB for student-led technical projects in the mechatronics domain.
- Collaborated with peers on small-team prototypes — covering data collection, model training, and basic validation for applied ML ideas.
- Supported chapter workshops and peer-learning sessions introducing ML fundamentals and hands-on coding exercises.
Projects
JanVaani — AI Scheme Assistant
FastAPI · PostgreSQL · Supabase · RAG · OpenAI · Vercel · Live
Designed a multilingual AI assistant for discovering government welfare schemes using Retrieval-Augmented Generation (RAG), semantic search, and vector embeddings.
- Built a FastAPI backend with Supabase and PostgreSQL for structured data and retrieval workflows.
- Implemented context-aware responses with low-latency semantic retrieval over scheme knowledge bases.
- Added user feedback collection to continuously improve answer quality and relevance.
- Deployed live on Vercel for accessible, production-ready citizen-facing access.
Geriatric Depression Support Chatbot
LLM · NLP · Sentiment Analysis · Twilio
Built a multimodal AI chatbot supporting voice and text interactions with multilingual NLP and real-time sentiment analysis. Integrated Twilio-based caregiver alerts for negative sentiment, reducing manual monitoring effort by approximately 95%.
Technical Skills
Education
Manipal Institute of Technology
B.Tech, Data Science Engineering
CGPA: 8.8 / 10.0
Jan 2020 – Present
Achievements
-
Finalist — Deriv × lablab.ai Hackathon
Built InsightX, an AI-powered trading assistant using LLMs, RAG, and financial analytics to deliver personalized trading insights and explainable decision support. -
Kaggle Notebook Expert
Recognized among Kaggle's top notebook contributors for publishing high-quality Machine Learning and Data Science notebooks.