DOC.AI
Production RAG system for PDF and DOCX ingestion. SQL-backed retrieval, multi-document comparison, and AssemblyAI voice interface. Built to handle real document workloads — not a chatbot wrapper.
1st Place · Need for Code 4.0
Yashas Vaddi
>_ AI/ML Engineer
I build AI that runs in production — not notebooks. Multi-agent pipelines, RAG systems, and cloud-deployed NLP services that handle real workloads.
scroll
↓
0
GPA
0
Hackathon Podiums
0
Published Papers
0
Production Deployments
Recognition
7 podiums. All under pressure.
Computer Engineering student at TSEC Mumbai (CGPA 9.01), currently building multi-agent clinical AI pipelines at Eka.Care. My work doesn't stop at a working demo — it goes through orchestration, fault tolerance, observability, and GCP deployment.
At TRREV Technology I shipped a FastAPI service that improved keyword extraction precision by 30–50% in production — not in a controlled eval, in a live system. That gap between "it works on my machine" and "it works under load" is exactly where I operate.
7 hackathon podiums. 2 published papers. Every project either deployed or competition-tested under pressure. I don't build things I can't ship.
Experience
Production, not sandbox.
Eka.Care
2025 — Present
Agentic AI Intern NOW
Bangalore, India
- Designing multi-agent clinical pipelines with tool-calling, memory management, and inter-agent communication layers.
- Building orchestration infrastructure for reliable LLM execution under real medical workloads — not synthetic benchmarks.
- Targeting production GCP deployment with structured fault tolerance, retries, and observability from day one.
TRREV Technology
2024 — 2025
AI/ML Intern
Mumbai, India
- Shipped a FastAPI-based NLP service on GCP Compute Engine that improved keyword extraction precision by 30–50% in a live product.
- Built and deployed retrieval and data processing pipelines integrated directly into the production backend — not a separate research track.
- Owned the full path from model selection to deployment, monitoring included.
Recognition
Wins that matter.
Hackathon Wins
Built fast. Judged harder.
Projects
Shipped work.
MindVoice
Multimodal AI system combining voice and chat interaction with research-grade evaluation metrics. PyTorch-based, GCP-deployed, built for touch-free accessibility under real usage conditions.
1st Place · Codessiance
ScriptChat
Agentic document workflow assistant using LangChain and vector retrieval. Structured reasoning over long-form scripts, Dockerized for deployment. Designed for repeatable, production-ready outputs.
1st Runner · TSEC Hacks
← drag to explore →
Technical Stack
What I actually ship with.
ML / DL
- PyTorch
- Transformers
- DistilRoBERTa
- TensorFlow
- Scikit-learn
NLP
- RAG Pipelines
- LLM Orchestration
- Information Retrieval
- Embeddings
- Sentiment Analysis
Backend
- FastAPI
- REST APIs
- Service Design
- Asynchronous Services
- TCP/IP Fundamentals
- Linux/OS Basics
Cloud / MLOps
- GCP Compute
- Docker
- Monitoring
- AWS (EC2, S3, Lambda, IAM)
- CI/CD Pipelines
Languages
- Python
- C++
- SQL
- JavaScript
Data
- Vector Search
- ETL Workflows
- Experimentation
- Pandas
- NumPy
- Feature Engineering
Contact
If you're building something real, let's talk.
Open to internships, research roles, and founding engineer conversations. I respond fast. If you've read this far, you already know I ship.