Overview
AI Engineer
Social Links
About
Chronically online. Probably on the spectrum. Professionally addicted to building software.
I’ve delivered software and AI solutions for government clients and SMEs, but honestly, I’d do it even if they didn’t pay me. Give me something to build, a laptop, and mild sleep deprivation and I’m entertained for weeks.
Stack
Python
TypeScript
JavaScript
Java
Go
SQL
Bash
React
Next.js
Tailwind CSS
HTML5
CSS3
Node.js
Express
FastAPI
GraphQL
gRPC
Microservices
Redis
REST API
WebSockets
PostgreSQL
MongoDB
MySQL
FAISS
pgvector
PyTorch
TensorFlow
scikit-learn
CNN / RNN
Transformers
OpenCV
PaddleOCR
Tesseract
YOLOv8
Leaflet.js
LangChain
LangGraph
RAG
Hugging Face
OpenAI API
Gemini API
Prompt Engineering
GCP
Docker
Kubernetes
GitHub Actions
Experience
Lom
Current Employer- Found automation opportunities across the business and built the tools for them.
- Built a content assistant for video editors: upload a raw podcast (1hr+), get back suggested chapter markers, B-roll cues, asset placement, and short-form Reels cuts from the same recording.
- Added title, description, and thumbnail copy generation to the same tool, plus flagging moments worth clipping for other platforms.
- Whisper
- LLM
- Structured Output
- React
- Transcription
- Automation
- Product Development
TechBharat AI CoE
- Built a road hazard detection and reporting platform for municipal authorities — citizens submit photos, the system classifies damage and routes it to the right department.
- Added client-side image validation with TensorFlow.js so bad uploads get rejected before hitting the server.
- Trained a CNN (ResNet-inspired) to classify road damage severity and assign a repair priority score.
- Built the admin dashboard in Next.js with Leaflet.js — shows a heatmap of reported issues and lets authorities track each one from report to resolution.
- Containerized with Docker, deployed on GCP Cloud Run, CI/CD via GitHub Actions.
- Python
- JavaScript
- TensorFlow.js
- CNN
- Next.js
- Leaflet.js
- Docker
- GCP Cloud Run
- GitHub Actions
Education
- Currently studying for a Bachelor of Science (BS) in Data Science.
- CGPA: 8.5 / 10.0.
- Rigorous training in Python, Data Structures, Algorithms, Machine Learning, Deep Learning, and Statistical modeling.
- Python
- Data Science
- Machine Learning
- Deep Learning
- Data Structures
- Algorithms
- Statistical Modeling
Projects(3)
Built for Vizag Municipal Government: scan a handwritten or printed form, and it auto-fills the corresponding fields on the employee's workstation.
- PaddleOCR + OpenCV pipeline extracts the data and maps it to the digital form schema.
- Served via FastAPI with a Next.js operator UI.
- In active discussion with Vizag Municipal Corporation for production deployment.
- PaddleOCR
- OpenCV
- FastAPI
- Next.js
- PostgreSQL
- Computer Vision
- OCR
Outbound sales agent built on LangGraph: finds leads from websites, researches the company, writes a personalised email, schedules follow-ups, and tracks the whole conversation — no human in the loop.
- Each sub-agent (lead finder, researcher, drafter, scheduler) runs independently.
- Shares state dynamically through LangGraph's state graph.
- LangGraph
- OpenAI
- Gemini
- PostgreSQL
- React
- FastAPI
- GenAI
- Multi-Agent
Support chatbot that reads company docs into a vector DB and answers customer questions via RAG.
- Low-confidence answers get flagged for a human agent instead of guessing.
- Resolved tickets get fed back into the knowledge base over time, so the answers improve as it's used.
- RAG
- Vector DB
- LangChain
- FastAPI
- OpenAI
- Gemini
- Vector Search