Current
Building SHAPI, self-maintaining APIs for teams and AI agents. Recent M.S. graduate in Artificial Intelligence from Rochester Institute of Technology, based in the San Francisco Bay Area.
AI engineer / researcher / builder
Agents, retrieval, evaluation and infrastructure. I build them, test them, break them and ship them.
Currently building SHAPI Self-maintaining APIs for teams and AI agents. View project →
01 / About me
Building SHAPI, self-maintaining APIs for teams and AI agents. Recent M.S. graduate in Artificial Intelligence from Rochester Institute of Technology, based in the San Francisco Bay Area.
Bachelor’s in Computer Science from R.M.K. Engineering College, then ML engineering at Novacis Digital and research at RIT, building document intelligence, semantic search and LLM pipelines.
AI agents, LLMs, RAG, evaluation, computer vision and cloud-native ML deployment, turning research into systems with measurable impact.
02 / Education
M.S. in Artificial Intelligence
Bachelor’s in Computer Science and Engineering
03 / Featured projects
Self-maintaining APIs for teams and AI agents.
Job search infrastructure built around a career graph.
Translate text, photos and documents into the language you understand best.
A personal agent that notices, proposes, and acts only after you approve.
A local-first trust layer for AI agents.
Can multiple AI agents detect when RAG systems hallucinate?
Semantic search across video scenes using natural language.
A control tower for tracing, debugging and auditing agent runs.
Real-time feature computation and multi-model serving.
04 / Work experience
RIT
Key achievements
Novacis Digital
Key achievements
05 / Technical skills
Pick a skill to see where I used it.
Languages
Python · used in RAGTruthAgentOpsChiefVideo SearchFeature PlatformDocument IntelligenceGraduate Researcher, RIT
TypeScript · used in SHAPINoriMotherTongueKerna
Rust · used in Kerna
LLMs & agents
Claude API · used in SHAPINoriKernaMachine Learning Engineer, Novacis Digital
OpenAI API · used in RAGTruthDocument IntelligenceMotherTongueMachine Learning Engineer, Novacis Digital
Model Context Protocol · used in SHAPINoriKerna
Multi-agent systems · used in RAGTruthDocument IntelligenceAgentOpsGraduate Researcher, RIT
Human-in-the-loop agents · used in ChiefKernaAgentOpsNori
RAG · used in RAGTruthDocument IntelligenceGraduate Researcher, RIT
LangChain · used in Document Intelligence
LangGraph · used in Document Intelligence
LLM evaluation · used in RAGTruthKernaGraduate Researcher, RITMachine Learning Engineer, Novacis Digital
Composio · used in Nori
Machine learning
PyTorch · used in RAGTruthVideo Search
Hugging Face Transformers · used in RAGTruthVideo SearchMotherTongue
LayoutLM / BERT · used in Machine Learning Engineer, Novacis Digital
CLIP · used in Video Search
OCR · used in Document IntelligenceMotherTongueMachine Learning Engineer, Novacis Digital
Scikit-learn · used in RAGTruthFeature Platform
Pandas & NumPy · used in RAGTruthFeature Platform
Search & data
Qdrant · used in Video SearchDocument Intelligence
FAISS · used in Machine Learning Engineer, Novacis Digital
pgvector · used in Nori
PostgreSQL · used in SHAPINoriAgentOps
SQLite · used in Kerna
Redis · used in Feature PlatformMotherTongue
Kafka · used in Feature Platform
Feast · used in Feature Platform
MLOps & cloud
Docker · used in SHAPINoriAgentOpsFeature PlatformVideo SearchDocument Intelligence
AWS SageMaker · used in Machine Learning Engineer, Novacis Digital
Kubernetes (EKS) · used in Machine Learning Engineer, Novacis Digital
Ray Serve · used in Feature Platform
MLflow · used in Feature Platform
Prometheus · used in Feature Platform
Backend & web
FastAPI · used in AgentOpsFeature PlatformVideo SearchDocument Intelligence
React · used in NoriMotherTongueAgentOpsVideo SearchKerna
Next.js · used in NoriMotherTongue
Tailwind CSS · used in NoriMotherTongue
Playwright · used in Nori
Static analysis (tree-sitter, ts-morph) · used in SHAPI
Testing (pytest, Vitest) · used in SHAPINoriAgentOpsFeature PlatformVideo SearchDocument Intelligence
Also familiar with
06 / Research
An empirical study of SE, ECA and a proposed Lite Channel Attention module across ResNet-18 and MobileNetV2 on CIFAR-10, measuring accuracy, parameters, FLOPs and latency.
A lightweight difficulty-aware training framework that learns and predicts hard examples with an auxiliary head, making training dynamics easier to interpret while preserving accuracy on MNIST.
A zero-shot multi-agent pipeline that decomposes RAG outputs into claims, verifies them against sources and flags unsupported spans, evaluated on RAGTruth.
07 / Certifications
NVIDIA
Edge AI deployment, model optimization and real-time inference.
Stanford University
Machine learning algorithms, theory and practical applications.
08 / Get in touch
Always open to discussing new opportunities, collaborations, or just having a chat about AI and ML.