CSE Grad · AI/LLM Engineer · RAG Systems · Agentic AI · Production ML
-
🔭 Working as AI Engineer Intern @ ASAR IT Technologies, Bangalore — production-grade RAG systems & deterministic LLM pipelines
-
🌱 Advancing in Agentic AI & Generative AI — real-world deployment, eval-driven dev & low-latency inference
-
💻 Currently building Pocket CEO · Agentic Medical Report Analysis · Conversational AI Avatar
-
🤝 Open to collaborating on open-source AI/LLM projects — retrieval systems, agentic workflows & scalable APIs
-
💬 Ask me about RAG, LLM pipelines, FastAPI, semantic search & production AI engineering
-
📫 Reach me at: irakamsivabhanuprakash@gmail.com
-
📱 Connect: LinkedIn/siva-venkata-bhanu-prakash
-
⚡ Fun fact: Consistent progress during tough times is the real compound interest.
AI / OCR Search Engineer Intern — ASAR IT Technologies, Bengaluru (Jan 2026 – Present)
- Built deterministic LLM extraction pipelines achieving 99% JSON-valid outputs at ~300ms latency
- Reduced OCR noise by 80%+ and improved extraction F1 score by 18%
- Implemented RAG pipelines using LangChain + FAISS with FastAPI-based inference & retrieval APIs
- Built evaluation dashboards tracking precision, recall, F1 & latency
| 🏆 MSME Agent-A-Thon — "Pocket CEO" | 🩺 Agentic Medical Report Analysis |
|---|---|
| Multi-agent LLM financial advisor for 100+ MSMEs · 60% less manual analysis · Shopify, Mailchat & Supabase APIs · Built at GDG Hyderabad Agent-A-Thon | Conversational medical-report RAG platform · FAISS-based semantic retrieval · Structured output via Pydantic · Agentic orchestration pipelines |
 Python LangChain LLMs Supabase React |
 RAG FAISS LangChain Pydantic Streamlit |
| 🤖 Conversational AI Avatar | 📰 News Analyzer |
|---|---|
| Multimodal AI avatar · Speech-to-Text & TTS · Agentic memory · Real-time interaction · FastAPI + React full-stack architecture | LLM-powered news fetching, analysis & summarization pipeline |
FastAPI React STT TTS Agentic Memory |
Python LLM NLP |
| 🩸 HematoVision | 💬 QA ChatBot (Open-Source Models) |
|---|---|
| Accurate blood cell classification using Transfer Learning & pre-trained CNNs · Efficient model for clinical-grade detection | Local LLM chatbot via Ollama · Runs 1B–8B parameter models on-device · Requires 8–16 GB RAM |
Transfer Learning CNN Jupyter Notebook OpenCV |
Ollama Python Open-Source LLMs |
|
🗣️ Languages
🤖 AI / ML / LLM
🌐 Backend & APIs
🗄️ Databases & Search
|
📊 Data Science
🎨 Frontend & Web
☁️ Cloud & Deployment
🛠️ Tools & DevOps
|
Real-time insights into my coding activity, languages, and contributions
Thanks for visiting! Feel free to explore my repositories and reach out for collaborations. 🤝