Anant Patankar · India · Open to ML/AI roles
ML/AI Engineer: Machine Learning to Agentic AI
Machine Learning · NLP & Time Series · Production RAG · Multi-agent Systems (Google ADK) · MCP · MLOps
Trained in forensic science, where every conclusion has to survive cross-examination. I build ML and AI systems the same way, from predictive models on messy real-world data to RAG and multi-agent systems grounded in evidence.
What I do
Machine Learning & Data Science
Forecasting, anomaly detection, classification and NLP models, from feature engineering on large datasets to evaluation and deployment, across banking and manufacturing use cases.
Agentic AI
Multi-agent orchestration with Google ADK, MCP servers, and LLM reasoning and tool use, in systems where the model makes judgment calls and the architecture keeps it safe, cheap, and auditable.
Production RAG
Grounded generation with source attribution, dense and hybrid retrieval, and cross-encoder reranking, evaluated with labelled retrieval metrics and latency monitoring.
Production Systems & MLOps
Async Python, distributed pipelines (Celery, Redis, Cloud Run), containerised deployments, experiment tracking and model registries: the backbone that makes ML and AI shippable.
Skills
Used in production work
Agentic AI & LLMs
Vector Search & Retrieval
Machine Learning & Deep Learning
Distributed Systems & Infra
MLOps & Cloud
Data Engineering
Familiar with
Selected work
Featured Projects
Five systems spanning agentic engines, MCP servers, and RAG pipelines. Two are open source and independently verifiable.
Multi-Agent Recommendation Engine for Gamified Data Labelling
Google ADK · Gemini 2.5 Flash · GCP · FlagshipBackend for a gamified platform with two goals: annotate an industrial safety image dataset, and raise frontline workers’ health, safety and environment awareness. One mode is a quiz that teaches; the others produce structured labels. 12 agents, 3 LLM-backed; the other 9 are deterministic Python handling scheduling, consensus, Elo ranking and trust scoring.
12 agents, 3 LLM-backed · per-skill tool gating · the model only at decision margins
Grounded Training Content Generation with Human Review
Vertex AI Agent Engine · RAG · Human-in-the-loop14 agents behind a single registry, shipped as one Vertex AI Agent Engine application that is the only ingress. Turns source documents into training questions, grounded per question with its own source attribution. Nothing reaches the live bank without explicit human approval.
multi-agent generation · multi-format RAG ingestion · no auto-publish path
Lumenore Analytics MCP Server
MCP Protocol · Async PythonMCP server that exposes a BI platform’s analytics to any MCP-compatible AI assistant: natural-language querying, trend, forecasting, correlation and outlier analysis over typed tools.
7 analytics tools plus metadata helpers · connection-pooled async client · Streamable HTTP
MLOps MCP Server
MCP Protocol · MLOps · Personal OSSOpen-source MCP server giving AI assistants tool-based access to MLOps workflows: experiment tracking, model registry, dataset drift detection, pipeline DAG validation and lineage tracing, wrapping MLflow, DVC and Git.
pip install mlops-mcp-server · v0.1.0 · MIT · 15 domain modules
Enterprise Knowledge Q&A
Production RAG · Retrieval · EvaluationProduction RAG over enterprise policy and procedure documents: dense retrieval with per-business-unit isolation through metadata filtering, and cited, grounded answers that refuse when retrieval confidence is low.
85%+ Recall@5 on a 200-query labelled set · per-business-unit isolation · cited answers
Verifiable proof
Open Source
Don’t take my word for it: read the code.
mlops-mcp-server
PyPIMCP server for MLOps workflows: experiment tracking, model registry, KS-test drift detection, DAG cycle detection (Kahn’s algorithm), and BFS lineage tracing with Mermaid visualisations.
lumenore-mcp
Org repoAnalytics MCP server I built at Netlink: FastMCP over Streamable HTTP, async aiohttp with connection pooling, and 7 analytics tools plus metadata helpers. Published open source under the Lumenore Platform organisation.
Track record
Experience
Sole AI engineer on two multi-agent systems on Google Cloud for an enterprise client: a recommendation engine for a gamified safety-labelling and awareness platform (12 agents, 3 LLM-backed) and a grounded content-generation platform with mandatory human review. Also mentored a junior engineer.
Built the open-source Lumenore Analytics MCP Server; an AI customer-support chatbot with hybrid retrieval (dense vectors plus BM25), cross-encoder reranking and asynchronous message processing on Celery and Redis; and an MCP client and orchestrator managing multiple MCP servers across HTTP, WebSocket, SSE and stdio transports, with role-based access control and parallel fan-out execution.
Built the retrieval layer of a production enterprise knowledge Q&A system (85%+ Recall@5 on a 200-query labelled set); an AI code-analysis system across 10+ programming languages; and data science use cases for a banking client spanning forecasting, NLP and OCR, with large-scale data processing on PySpark. 2× Game Changer of the Month (Sept and Dec 2023).
Built Random Forest and SVM models for reactor-parameter and yield optimisation, and an LSTM autoencoder for real-time anomaly detection.
Worked on process data analysis and model prototyping for manufacturing optimisation.
Built an ML forecasting pipeline (ARIMA, Random Forest, LSTM) and a real-time object-detection system for video feeds with TensorFlow.
About
From forensic science to machine learning and AI
I hold an M.Sc. in Forensic Science and moved into machine learning through self-directed study. I have been building ML systems since 2019 and Generative AI systems since 2023. The analytical discipline carried over: whether it is a predictive model or a multi-agent system, I care about how it is validated, what evidence backs its output, and whether its decisions can be audited.
My work spans classical machine learning and Generative AI: forecasting, anomaly detection and NLP models on one side, production RAG and multi-agent systems with Google ADK and MCP on the other, supported by distributed systems and MLOps practice. Two open-source MCP servers I built are publicly available.
Get in touch
Contact
Currently open to ML/AI Engineer roles and available to join immediately. Happy to talk machine learning, Generative AI and agentic systems.