Software Engineer (AI/ML) building agentic AI platforms in production. 6+ years across software engineering, ML research, and enterprise automation. Currently at Bayesics, where I architect a reusable multi-agent framework powering internal AI applications for NASA program teams and federal R&D partners. MS Computer Science from Rowan University (4.0 GPA).
- Agentic AI platforms. Multi-agent orchestration with LangGraph, sandboxed tool execution, LLM-as-judge evaluation pipelines, prompt-injection defenses, and context-engineering primitives for production deployment.
- Open-source MCP servers. Domain-specific Model Context Protocol servers that connect AI assistants to real-world data and tools. Three so far - legal research, construction trades, and HR compliance.
- LLM safety and evaluation. Fine-tuning, machine unlearning, grounding judges, and golden-dataset regression as part of any AI shipping discipline.
| Repo | What it is |
|---|---|
| legal-mcp | Comprehensive US legal MCP server. 18 tools spanning case law (4M+ court opinions via CourtListener), practice management (Clio), and federal filings (PACER). MIT licensed. |
| trades-mcp | First MCP server for construction trades. 13 tools for contractor license verification across CA/TX/FL/NY, BLS labor rates, material pricing, project cost estimation. 68+ passing tests. MIT licensed. |
| Machine-Unlearning-SLM | Fine-tuned Llama 3.2 3B on 12K+ synthetic records to filter 500+ protected entities. Multi-GPU inference on dual NVIDIA P40s, 2.3s -> 1.1s latency on 10K query evaluations. |
| GNN-for-forecasting-coordinated-moving-objects | Spatio-temporal GCN-LSTM models for multi-agent trajectory forecasting. 22-node temporal graphs, 300K+ frames. From my MS research at Rowan. |
AI / agents. LangGraph, AutoGen, MCP (Model Context Protocol), Anthropic Claude, OpenAI GPT, Ollama, LangChain, Hugging Face (Transformers, PEFT), pgvector, FAISS, RAG, LLM-as-judge.
Backend. Python 3.11 (asyncio), FastAPI, SQLAlchemy 2.0, PostgreSQL, Redis, Celery, Pydantic v2.
Cloud and infra. AWS (EKS, RDS, Lambda, S3, Bedrock, SageMaker), Azure (AKS, AI Foundry), Docker, Kubernetes, Terraform, GitHub Actions.
Practices. TDD, hexagonal architecture, evaluation-driven shipping, OWASP LLM Top 10 + NIST AI RMF security baselines.
- Email: mahendersam5640@gmail.com
- LinkedIn: linkedin.com/in/mahender-samperboyina-43a865142
- Currently open to: Software Engineer / AI Engineer roles, US-based, hybrid or remote
Some of my domain-specific MCPs draw from public datasets and APIs (CourtListener, PACER, Clio, CSLB, BLS, NYC Open Data, Texas open data, MyFloridaLicense). Thanks to those teams for keeping the data accessible.