BUILD.LEARN.CAPTURE.
A few things I've built end to end.
RecoverRL
FAILURE → FEATURES → LEARNED POLICY → VERIFY
Agents that learn how to recover: when a tool call or multi-step workflow fails, a learned policy picks the right recovery strategy instead of blindly retrying.
ReMem
ADD → SEARCH → ACT → FEEDBACK → LEARN
Memory for AI agents that learns from outcomes which memories actually help — it retires outdated facts, demotes useless look-alikes, and fits a token budget.
PaperForge AI
RESEARCH → RETRIEVAL → CODE
An end-to-end RAG pipeline that reads academic research papers and turns their methodology sections into working Python.
The other half of how I look at things.
Race days, campus mornings, skies that showed off — shot first, sorted later.
DRAG TO LOOK ALONG THE WALL · CLICK A PRINT TO OPEN IT
Wipro
AI/ML Intern
MAY 2026 — JULY 2026
Hyderabad, India
Built a RAG pipeline over messy, real-world domain PDFs where answer accuracy mattered, and developed an agentic workflow for a client's process.
- RAG over 25+ real-world PDFs: LangChain, Hugging Face embeddings, ChromaDB
- Hallucination control through chunking, Top-K, and prompt validation
- Agentic workflow with Google ADK to optimize a client's process
lab note
RAG PIPELINE
25+
PDF documents
1,000+
indexed chunks
- 25+ DOCS
- 1,000+ CHUNKS
- EMBEDDINGS
- CHROMADB
- RETRIEVAL
- LLM
- ANSWER
An actual arcade lives in here.
Six classics rebuilt from scratch and genuinely playable — keyboard on desktop, touch controls on mobile.
- PYTHON
- RAG
- LLMs
- LANGCHAIN
- LANGGRAPH
- PYTORCH
- FASTAPI
- DJANGO
- DOCKER
- FAISS
- HUGGING FACE
- MONGODB
Systems, stories, speed.
Hi, I'm Viswesh Kesarla — a Computer Science student who ends up spending most of my time in the gap between an idea and a system that actually runs.
I enjoy building with LLMs, RAG, backend technologies and emerging developer tools. Right now that means retrieval pipelines, evaluation, and getting language models to output things that compile.
Outside of code, I'm usually taking photographs or watching a race — the same instinct, really: paying attention to timing, framing, and what happens right at the edge.
CURRENTLY EXPLORING · OPEN TO INTERNSHIPS & COLLABORATIONS
education
Woxsen University
B.Tech in Computer Science and Engineering
2024—2028
DURATION
8.61 / 10
GPA
RELEVANT COURSEWORK
- Data Structures & Algorithms
- Object-Oriented Programming
- Database Management Systems
- Computer Networks
- AI/ML Foundations
LET'S
BUILD
SOMETHING.
visweshkesarla@gmail.com