Applied AI / LLM Engineer building evaluated agent systems, RAG pipelines, and production AI tooling.
I build production-minded AI systems with Python, LangGraph, LangChain, FastAPI, OpenAI, Anthropic, Streamlit, SQLite, and vector search. My work focuses on retrieval reliability, agent memory, evaluation, observability, privacy boundaries, and cost-aware execution.
I am a 2026 B.Tech graduate and AI Engineering Fellow at Maven (AI Makerspace), open to Applied AI / LLM Engineer roles with Indian and international AI teams.
- Built recruiter-ready AI systems across coding agents, RAG, data generation, privacy, memory, and context engineering.
- Engineered DevMind with six security-aware tools, persistent sessions, runtime metrics, plugin support, CI, and 156 tests.
- Built ContextOps Agent with typed memory, plan persistence, context compression, privacy review, evaluation metrics, and observed CI success.
- Built OpenAI AutoData with challenger/solver/judge agents, persistent budget controls, fail-closed validation, auditable outputs, 13 regression tests, and CI.
- Built research-backed RAG systems with corrective retrieval, adaptive routing, citation grounding, and RAG evaluation metrics.
- Completed Andrew Ng's five-course Deep Learning Specialization and continue studying production RAG, agent evaluation, and context engineering.
| Project | What it does | Engineering signals |
|---|---|---|
| DevMind | Terminal-native AI coding agent built with Python, LangGraph, and Claude | Six built-in tools, persistent sessions, runtime metrics, plugins, cross-platform support, 156 tests, CI |
| ContextOps Agent | Context-engineering layer for long-horizon agents with typed memory, compression, and privacy review | Plan persistence, memory graph reconstruction, privacy firewall, token-savings metrics, FastAPI, Streamlit, CI |
| Secure RepoPilot | Issue-to-PR coding agent with baseline verification, safety controls, and privacy auditing | Reproducer, minimal patching, command guardrails, leakage audit, SWE-style judge, FastAPI, Streamlit, CI |
| OpenAI AutoData | Generates hard research QA data through challenger, solver, and judge agents | Persistent budget guard, fail-closed validation, auditable outputs, 13 regression tests, CI |
| Corrective Agentic RAG Assistant | Adaptive CRAG assistant that detects retrieval failure and corrects noisy context | Query router, corrective retrieval actions, hierarchical retrieval, RAG metrics, CI |
| TrustDI Agentic RAG | Agentic RAG system for trustworthy enterprise data integration and schema matching | Adaptive routing, evidence-backed decisions, OpenAI explanations, precision/recall/F1 evals, FastAPI, Streamlit, CI |
- Long-horizon agents with explicit memory, context budgets, and quality gates
- RAG systems with retrieval evaluation, citation grounding, and failure recovery
- Tool-using agents with privacy boundaries, observability, and cost controls
- AI evaluation for faithfulness, reliability, safety, and regression testing
- Python APIs and deployable AI applications
- MemoryOS Agent: MemGPT-inspired long-term memory agent with OpenAI API support, SQLite memory, lifecycle controls, and Streamlit dashboard.
- Adaptive RAG CAG Project: adaptive RAG and cache-augmented generation demo for retrieval workflows.
- Local RAG with Ollama and ChromaDB: private PDF question answering with local inference and embeddings.
- AI Reddit Brand Monitor: local sentiment, topic, urgency, and feedback analysis for Reddit mentions.
Python LangGraph LangChain FastAPI OpenAI API Anthropic API Streamlit ChromaDB SQLite PyTorch Docker GitHub Actions RAGAS-style evals
- Build the smallest reliable system that proves the idea.
- Test failure paths, not only happy paths.
- Make cost, state, and model behavior visible.
- Keep claims aligned with reproducible code and results.
- Document limitations clearly instead of overstating benchmark performance.
I am open to Applied AI / LLM roles and focused open-source collaboration. If you find a project useful, follow the profile for upcoming builds or star the repository you want to revisit. Technical feedback is welcome.
Based in India. The strongest repositories are pinned below for a quick technical review.