$ python finetune_lora.py
↳ Indexing 10K docs → FAISS… done
↳ Retrieval p95: 384ms
$ sagemaker deploy ––env prod
↳ Endpoint: InService ✓
$ |
AI/ML Engineer with around five years of experience building production LLM systems, Generative AI applications, and scalable distributed backend infrastructure. Hands-on across the full stack — RAG pipelines, model fine-tuning, agentic AI, microservices architecture, and cloud-native deployments on AWS. Most recently at Piper Sandler, where I built a LangChain + FAISS retrieval layer over 10K+ SEC filings and earnings transcripts that cut analyst lookup time from ~8 seconds to under 400ms, and shipped a Generative AI portfolio assistant on the OpenAI and Claude APIs. Previously at Dell Technologies, I overhauled the Java Spring Boot + Kubernetes microservices layer behind the OpenManage server platform (1,200 → 8,000 concurrent sessions) and migrated firmware telemetry to Kafka event streaming for 600K+ endpoints; at Airbnb, I shipped Python/Django REST APIs and Redis caching for host availability and messaging. That production work carries into my own projects — Kube-LoRA, a Kubernetes-native LoRA fine-tuning platform; a monitored ChestX-ray14 classification platform with three-layer drift detection; a CLIP + XGBoost learning-to-rank engine; and an autonomous ReAct research agent on LangGraph. Currently completing an MS in Applied Machine Intelligence at Northeastern University in Boston.
Northeastern University | Boston, MA, USA | Sep 2024 - Apr 2026
Master of Science in Applied Machine Intelligence, focused on deep learning, MLOps, and applied LLM systems.
Kristu Jayanti College | Bangalore, India | Jun 2019 - May 2022
Bachelor of Science in Computer Science, completed alongside professional software engineering work.
St. Philomena's Public School & Junior College | 2017 - 2019
Successfully completed high school in the concentration Computer Science
Piper Sandler | Boston, MA, USA | Jun 2025 – Nov 2025
Dell Technologies | India | Aug 2022 – Aug 2024
Airbnb | India | May 2020 – Jul 2022
IIIT Bangalore - Advanced Certification Program for Data Science
LinkedIn Learning - Azure Administration Essential Training
AWS Academy Cloud Foundations - Comprehensive cloud computing fundamentals
Anthropic - Foundational certification in designing and architecting solutions with Claude
Build production assistants and copilots on the OpenAI, Claude, and Bedrock APIs — function calling, structured outputs, tiered model routing, and evaluation with LLM-as-judge.
Design retrieval pipelines over large document sets with LangChain, FAISS, and vector databases, tuned for sub-second lookup and grounded, citable answers.
Ship autonomous agents on LangGraph that plan, call tools, and iterate on their own observations, with Pydantic-validated outputs and bounded recursion.
Fine-tune and train models with LoRA/QLoRA and PEFT in PyTorch under mixed precision and memory-constrained GPU budgets, from experiment tracking to quality gates.
Deploy and monitor models on AWS SageMaker, EKS, and Azure ML with Docker, Kubernetes, Terraform, MLflow, and blue-green canary promotion.
Architect scalable microservices and event-driven platforms with Spring Boot, FastAPI, Kafka, and Redis — built for high throughput and low latency.
Asynchronous microservices pipeline using FastAPI and Redis for speech recognition (Whisper) and sentiment analysis (RoBERTa), with Docker containerization and horizontal scaling.
Kubernetes-native fine-tuning platform driving training, scoring, and inference from a 14-endpoint FastAPI gateway with scoped RBAC. Rank-32 LoRA fine-tuning of Stable Diffusion v1.5 fits a 10.8GB budget on a 12GB RTX 3060 via fp16, gradient checkpointing, SDPA, and 8-bit Adam, with Celery/Redis workers and an automated ResNet50 quality gate feeding a scale-to-zero inference server.
End-to-end MLOps platform for multi-label pathology detection on NIH ChestX-ray14 (~112K images) using DenseNet-121, Azure ML, Terraform IaC, and blue-green canary deployment with GradCAM/SHAP explainability.
Autonomous ReAct agent on LangGraph that plans, calls tools, and iterates on its own observations, returning Pydantic-validated synthesis with a testable hypothesis. Tiered routing runs the loop on Claude Haiku and reserves Claude Sonnet for final synthesis, over an arXiv + FAISS knowledge base extended with live search via MCP.
Fault-tolerant distributed scheduling system with leader election, heartbeat monitoring, crash recovery, and Prometheus/Grafana observability.
Production ML system with fine-tuned DistilBERT for sentiment classification, LangGraph-based intelligent agent routing, and SageMaker deployment.
Cloud-native ETL pipeline processing FDA drug approvals and clinical trial data using Airflow DAGs, KubernetesPodOperator, AWS S3 data lake, and GitHub Actions CI/CD.
RAG-based conversational agent for automating academic advisor workflows, built with LangChain, Supabase vector databases, and real-time query processing.
Real-time AR application with MediaPipe Face Mesh, Three.js, achieving 30+ FPS on web and mobile platforms.
AI-powered desktop monitoring with facial recognition, emotion detection, and real-time productivity analytics dashboard.
Full-stack responsive web app for gaming community with Firebase real-time sync and authentication.
Interactive visualizations and comprehensive data analysis for loan application patterns and insights.
End-to-end DuckDB ETL pipeline processing 30GB+ transactional data, with A/B test simulation validating a 4.12% LTV lift, Streamlit visualization, and Power BI executive dashboards.
AI-powered multimodal coaching platform with NestJS microservices, MediaPipe video analysis, multi-provider LLM router, RAG with pgvector, and Expo mobile app with offline-first architecture.
End-to-end recommendation engine with CLIP embeddings, XGBoost Learning-to-Rank, PySpark data processing, Redis feature store, Airflow orchestration, and Evidently AI drift monitoring.