Open to GenAI roles · Bengaluru, India
I build production-grade GenAI systems — RAG pipelines, LLM apps, and the MLOps rails that keep them running.
The Story
Most people pick one side — research or infrastructure. I deliberately learned both, because the models that matter are the ones that ship.
St. Joseph Engineering College, Mangalore
Started a B.E. in Artificial Intelligence & Machine Learning. What began as coursework became obsession — training first neural networks, losing weekends to Kaggle, and realising models in notebooks are only half the story.
Future Interns · Elevate Labs
Internships flipped the perspective from 'can I train it?' to 'can users rely on it?'. Built and deployed real ML features, learned CI/CD discipline, and discovered that the gap between a demo and a product is engineering.
MATRIQZ AI, Bengaluru — DevOps Engineer Trainee
Went deep on the rails AI runs on: Docker, Kubernetes, AWS, Jenkins pipelines. Deploying other people's models taught me exactly what breaks in production — knowledge most model-builders never get.
LLM apps · RAG systems · MLOps
Combining both halves: designing RAG architectures with LangChain and Qdrant, then shipping them behind FastAPI on cloud-native infrastructure. From prompt to pod, I own the whole loop.
Capabilities
Three layers, one loop: generative systems on top, ML and APIs in the middle, cloud-native infrastructure underneath.
Retrieval, orchestration, and prompt systems for production LLM apps.
Model training and the APIs that serve them.
The rails that keep models alive in production.
Track Record
MATRIQZ AIBengaluru
Future InternsRemote
Elevate LabsRemote
Selected Work
Each one covers the full loop — a real problem, a working system, and the architecture that keeps it running.
RAG-powered repair assistant
ProblemDevice repair knowledge is scattered across thousands of unstructured guides — impossible to search when your phone is in pieces on the table.
SolutionA retrieval-augmented chatbot that ingests repair manuals into a vector store and answers step-by-step repair questions with cited sources, grounded in the actual guides.
ArchitectureDocs → chunking + embeddings → Qdrant vector store → LangChain retriever → GPT-4o generation with source citations, served over FastAPI.
Generative brand identity engine
ProblemEarly-stage founders need logo and brand assets fast, but designers are expensive and generic AI images ignore brand consistency.
SolutionA fine-tuned SDXL LoRA pipeline that generates on-brand logo concepts from a text brief, with style controls for palette, mood, and industry.
ArchitectureBrief → prompt templating → SDXL + custom LoRA weights → post-processing & upscaling → asset variants, orchestrated behind a FastAPI service.
AgriTech marketplace intelligence
ProblemFarmers sell at whatever price the nearest middleman quotes — with zero visibility into real mandi (market) rates across regions.
SolutionAn AgriTech platform surfacing live crop prices across markets with ML-backed price trends, helping farmers decide where and when to sell.
ArchitectureMarket data ingestion → cleaning + trend models → REST API → responsive web client with regional language support.
Kubernetes-deployed NLP service
ProblemSentiment models usually die in notebooks — real products need them autoscaling behind an API with zero-downtime deploys.
SolutionA sentiment analysis microservice packaged as a cloud-native app: containerised, orchestrated on Kubernetes, and continuously delivered through a CI/CD pipeline.
ArchitectureTransformer sentiment model → FastAPI inference service → Docker image → Kubernetes deployment on AWS with Jenkins CI/CD and health probes.
Real-time computer vision
ProblemTouchless interfaces need gesture input that works in real time on commodity hardware — no depth cameras, no lag.
SolutionA real-time hand gesture recognition system using landmark detection and a lightweight classifier, mapping gestures to actions at webcam frame rates.
ArchitectureWebcam stream → hand landmark extraction → keypoint features → CNN classifier → action mapping, optimised for real-time inference.
Proof of Work
Cloud fundamentals, compute, storage, and deployment on AWS.
Hands-on skill badges across GCP compute and ML services.
Andrew Ng's ML specialization — supervised, unsupervised & best practices.
Industry job simulations in software engineering and data.
Always Building
Pulled live from the GitHub API — no vanity widgets, just the actual data.
Across public repositories
Let's Talk
I'm open to GenAI / LLM engineering roles and interesting collaborations. The fastest way to reach me is email — I reply within a day.