Research Assistant, Indiana University School of Optometry
- Built automated point-cloud analysis pipelines in Python + Open3D.
- Improved 3D terrain reconstruction throughput by 25% with optimized preprocessing and meshing.
Actively Interviewing | US AI/ML Roles
Sarabesh Neelamegham Ravindranath
AI/ML Engineer with 4+ years across MLOps, DevOps, computer vision, and retrieval systems. I build production-ready ML pipelines and ship reliable models from experimentation to deployment.
AI/ML Engineerproduction ml systems
MLOps Engineertraining, deploy, monitor
Applied Scientistvision + nlp research to prod
LLM/RAG Engineerretrieval, reranking, eval
Recruiter Snapshot
- Target Roles: AI/ML Engineer, MLOps Engineer, Applied Scientist, LLM/RAG Engineer
- Education: MS Data Science, Indiana University Bloomington (GPA 3.90/4.00)
- Location: Bloomington, Indiana | Open to relocation within the US
Work Authorization Notice: F-1 OPT valid through June 2026. Eligible for 24-month STEM OPT extension. Requires H-1B sponsorship thereafter (Change of Status).
Experience
Recent Roles
Emerging Tech Intern (AI Specialist), Musco Lighting
- Created modular MLOps pipeline with DVC and GitLab CI for retraining automation.
- Integrated YOLOv5 training and inference with MLflow tracking and model observability.
DevOps Engineer, MTS II, VMware
- Shipped multi-cloud install and validation automation for Tanzu Application Platform.
- Expanded Go-based E2E test coverage from 20% to 75%.
Cloud Engineer, Presidio
- Built enterprise CI/CD and IaC solutions across AWS, Terraform, and Kubernetes.
- Reduced build times by 35% by migrating Jenkins workers to Kubernetes pods.
Projects
Selected Projects for AI/ML Roles
RAG
HybridRAG
Production-style hybrid retrieval architecture combining vector and graph search for higher-precision LLM grounding.
View RepositoryVDB
PuppyDB
Custom vector database built with LMDB persistence and ANN indexing to understand high-performance embedding retrieval.
View RepositoryPPO
GridWorld-PPO
Reinforcement learning path-planning with PPO in dynamic grid environments, including reward shaping and policy evaluation.
View RepositoryCV
Neural-Recon
fMRI-to-image reconstruction pipeline using diffusion models and CLIP-aligned features for vision-neuroscience experimentation.
View RepositoryRERANK
Retrieval-Reranker
Retrieval and reranking benchmarking across BEIR-style workflows to improve relevance and downstream answer quality.
View RepositoryRECSYS
reddit-recsys
Multimodal recommendation system using CLIP embeddings and vector search with workflow orchestration for real-time retrieval.
View RepositorySFT/RLHF
Finetuning
Baseline repository for LLM post-training (SFT/RLHF-style) and evaluation workflows with reproducible experimentation.
View RepositoryTRANSFORMER
exploring-transformers
Hands-on transformer architecture implementation with tokenizer, positional encoding, and training-focused experiments.
View RepositoryDEEPVO
mono-cam-DeepVO
Reimplementation of DeepVO for monocular camera motion estimation and robust trajectory prediction in vision pipelines.
View RepositoryBERT OPS
sentiment-analysis-finetuning-and-deployment
End-to-end BERT fine-tuning, API deployment, monitoring, and automated retraining workflow for production MLOps simulation.
View RepositoryNER
SkinNER
Clinical NER pipeline for skin-disease tagging using spaCy and BERT fine-tuning with comparative model evaluation.
View RepositoryVIT
Algae-Classification
Computer vision classification workflow for algae detection and water-quality use cases using deep learning approaches.
View RepositoryBlogs
Writing & Technical Notes
Vector Databases
How I Built a Vector DB with HNSW from Scratch
Building PuppyDB from scratch using HNSW and ANN search fundamentals.
Read on MediumGenerative AI
Image Reconstruction from Brain Signals
Reconstructing visuals from fMRI signals using ridge regression and diffusion models.
Read on MediumLLMs
1.58 BitNet
How BitNet-style quantization can reduce memory while preserving strong LLM performance.
Read on MediumSkills
Technical Stack
ML / AI
PyTorch, TensorFlow, Scikit-learn, Hugging Face, LangChain, Open3D, TensorRT, YOLOv5
MLOps / Infra
DVC, MLflow, SageMaker, Airflow, Prefect, Argo CD, Docker, Kubernetes, Terraform
Cloud / DevOps
AWS (Lambda, ECS, EKS, RDS, S3, Step Functions), GitHub Actions, GitLab CI/CD, Jenkins, Helm, Tanzu
Languages
Python, Go, C++, Java, Groovy, Flask, Spring Boot
Databases
MySQL, DynamoDB, Neo4j, Qdrant, PuppyDB
AI Domains
Computer Vision, NLP, Reinforcement Learning, Retrieval-Augmented Generation (RAG)
Education & Certifications
Credentials
MS in Data Science, Indiana University Bloomington (Aug 2023 - May 2025), GPA 3.90/4.00
BE in Computer Science, Anna University (Jul 2015 - Apr 2019), GPA 8.39/10.00
Certifications: AWS SAA, CKA, CKS, HashiCorp Terraform Associate
Work Authorization (US): F-1 OPT valid through June 2026. Eligible for 24-month STEM OPT extension. Requires H-1B sponsorship thereafter (Change of Status).
Let's Build Production AI Systems
Open to full-time AI/ML and MLOps roles in the US. Available to interview immediately.