PrototypeProduction
01Replace one-off demos with guarded services, retries, observability hooks, and clear deployment boundaries.
Available for AI/GenAI projects
AI Systems Engineer for the Agentic Era
Hi, I’m Prathmesh Adsod — a Python Developer and Generative AI Engineer building multi-agent systems, RAG pipelines, MCP tools, and production-grade AI automation.
Systems transition
From experiments to production-grade AI systems.
PromptAgentic Workflow
02Move from single prompts to stateful graphs with routing, validation, tool execution, and human approval gates.
Document SearchRAG Intelligence
03Turn static knowledge into retrieval systems with embeddings, hybrid search, reranking, citations, and evaluation loops.
2+
Years experience
Enterprise Python and GenAI delivery
700+
LeetCode problems
Algorithmic consistency under pressure
30+
Hackathon entries
Fast product thinking and demo execution
GenAI
Python + full-stack builder
LLM workflows, APIs, UI, deployment
TCS
Enterprise project work
Supply-chain automation context
BLR
Bangalore, India
Available for AI product builds
Signature expertise
Where I operate best
The strongest part of the portfolio is not a list of frameworks. It is the ability to compose agents, retrieval, tools, APIs, and product interfaces into systems that can be operated.
01
Multi-Agent Systems
LangGraph workflows, conditional routing, state management, tool-calling nodes, validation agents, execution agents, and agent-to-agent coordination.
LangGraphRoutingStateA2AValidationTools
Router
Planner
Retriever
Validator
Executor
02
RAG & Knowledge Systems
Chunking, embeddings, hybrid search, reranking, retrieval evaluation, latency tuning, and production-grade knowledge assistants.
ChunkingEmbeddingsHybrid SearchRerankingEval
Docs
Index
Retrieve
Rerank
Answer
03
MCP / Tooling Infrastructure
FastMCP servers exposing tools, resources, prompts, and structured context to LLM agents.
FastMCPResourcesPromptsTool ServersSchema
LLM
MCP
Tools
Resources
Context
04
Full-Stack AI Products
Next.js, React, Node.js, FastAPI, PostgreSQL, Firebase, GCP, Docker, APIs, and polished product UX.
Next.jsFastAPIPostgreSQLDockerGCPUX
UI
API
LLM
DB
Deploy
Experience
Enterprise AI experience
Production-style GenAI work in an enterprise supply-chain domain, with agent workflows, RAG, MCP tooling, integrations, and reliability patterns.
Oct 2023 – Present/Bangalore, India
Tata Consultancy Services
Python Developer – Generative AI
RSI Supply Chain Project
Architected multi-agent supply-chain automation workflows using LangChain and LangGraph.
Designed agent graphs with conditional routing, state management, and tool-calling nodes.
Built FastMCP servers exposing supply-chain tools, resources, and prompts to LLM agents.
Implemented A2A communication between planning, retrieval, validation, and execution agents.
Developed RAG pipelines for internal knowledge retrieval with chunking, embeddings, indexing, hybrid search, and reranking.
Integrated LLM inference with external APIs and internal microservices.
Added reliability patterns like error handling, retry logic, observability hooks, and documentation.
AI systems cockpitLive workflow
1
Documents
2
Embeddings
3
Retriever
4
Agents
5
Tools
6
Business Action
Featured case studies
Product stories, not project tiles
These are framed like AI products because that is how clients, founders, and judges evaluate serious systems: problem, solution, architecture, stack, and why it matters.
Case study 01Gemma 4
Presence
On-device AI persona platform
An on-device AI persona system that fine-tunes Gemma 4 using Unsloth LoRA on personal data, integrates Kokoro TTS and faster-whisper for real-time voice I/O, and uses RAG for persistent long-term memory.
Gemma 4Unsloth LoRAKokoro TTSfaster-whisperRAG
Local persona runtime
Gemma 4
LoRA
Memory
Persistent memory graph
factsvoiceeventspreferencescontextstylelong-termprivate
Case study 02LangGraph
Ghost
Agentic digital footprint cleanup system
An AI agent system that audits and removes a user’s digital footprint across services using Auth0 Token Vault for secure credentials and CIBA for human-in-the-loop approval on sensitive actions.
LangGraphAuth0 Token VaultCIBAPython
Privacy action map
1Audit services
2Classify risk
3Request approval
4Execute cleanup
Sensitive action requires approval
HoldApprove
Case study 03LangGraph
ShipSafe
Adversarial multi-agent merge request review
An adversarial MR code-review system where independent critic and defender agents surface security vulnerabilities and design issues beyond single-pass static analysis.
LangGraphGitLab APIPython
MR review debate
GitLab MR #42security/payment-flow.ts
- const token = req.headers.authorization+ const token = await vault.resolve(req)+ audit.trace("payment_guard")
Critic agent
Flags auth, data leakage, and regression risk.
Defender agent
Tests evidence and removes low-confidence noise.
High-confidence review packet generated
Case study 04Node.js
StockPick AI
Real-time AI financial analysis assistant
An intelligent financial advisor using Perplexity Sonar API for real-time investment analysis with caching, rate limiting, and Firebase deployment.
Node.jsExpress.jsReact.jsPerplexity Sonar APIFirebase
Market intelligence console
AI insight panel
Real-time research, cached context, rate limits, and deployment-aware analysis.
cachesourcesrisk
@PrathmeshAdsod
Build history, not just claims
68
Public repos
0
Profile signal
The main page keeps the strongest AI systems forward. Older learning repos and forks stay out of the primary proof layer.
Open GitHub profileGitHub intelligence
Public work, curated for relevance.
Python
gitlab-merge-request-smart-AI-plugin
GitLab merge-request AI plugin focused on intelligent review assistance.
0Jun 2025Featured AI
Repository
csb-incident-remedy-dataset
Production-grade dataset pipeline built from public CSB incident material for structured remediation intelligence.
0May 2026Featured AI
TypeScript
ShipSafe
Adversarial multi-agent merge request review system built around practical code-review workflows.
0Mar 2026Featured AI
JavaScript
conatorAI
Recent public repository from Prathmesh Adsod's GitHub build history.
0Jun 2025Featured AI
How I build
My engineering operating system
01
Understand business problem
Start from workflow pain, risk, users, and measurable product outcomes.
02
Design agent architecture
Map planner, retriever, validator, executor, memory, and approval boundaries.
03
Build RAG/tooling layer
Create retrieval, MCP tools, API connectors, schemas, and context contracts.
04
Add evaluation + observability
Track reliability, retry paths, retrieval quality, latency, and failure modes.
05
Ship product-quality UI
Turn backend intelligence into a polished interface people can actually operate.
06
Iterate with real feedback
Use traces, user feedback, and edge cases to strengthen the system.
LangGraphFastMCPA2A ProtocolRAGvLLMPyTorchFastAPINext.jsPostgreSQLDockerGCP
Client paths
Ways I can help
Clear, buildable engagements for AI products, retrieval systems, prototypes, and automation workflows.
01
AI Agent MVP
For founders who want to build agentic products fast.
Outcome: A working agent workflow with product UI and deployment-ready structure.
LangGraph / agent workflow
Tool calling
API integration
Dashboard UI
Deployment-ready structure
LangGraphFastAPINext.jsTool calling
Discuss project02
RAG Knowledge Assistant
For companies with documents, SOPs, internal knowledge, or support content.
Outcome: A retrieval system that can answer with grounded context and citations.
Document ingestion
Embeddings
Hybrid search
Citations
Evaluation-ready architecture
EmbeddingsHybrid searchRerankingEvaluation
Discuss project03
GenAI Product Prototype
For hackathons, demos, and startup MVPs.
Outcome: A polished prototype that explains itself through product experience.
Next.js frontend
Python/FastAPI backend
LLM integration
Polished UI/UX
Demo storytelling
Next.jsPythonLLM APIsFirebase
Discuss project04
AI Automation Workflow
For repetitive business processes.
Outcome: A controlled automation flow with approvals, retries, and visibility.
Multi-step agent flow
Human approval checkpoints
Retry/error handling
Observability hooks
AgentsA2AObservabilityAPIs
Discuss projectAI command center
A portfolio interface for serious AI work.
The quick scan: what I build, where I am strongest, and how to contact me when the project needs more than a prompt demo.
Prathmesh.system
Agentic portfolio runtime
Current focus
Multi-agent systems, RAG, MCP, AI automation
Strongest stack
Python, LangGraph, FastAPI, React, Node.js, PostgreSQL
Proof
700+ LeetCode, 30+ hackathons
Open to
AI product builds, GenAI systems, full-stack AI projects
Credibility
Built for serious AI product work
No invented testimonials. Just the signals that matter for someone evaluating whether Prathmesh can build real AI systems.
Enterprise project delivery
Works inside production-style supply-chain GenAI contexts with reliability, documentation, and integration constraints.
Hackathon product builder
Comfortable turning ambiguous ideas into demos with clear architecture, UX, and technical narrative.
Competitive programming consistency
700+ LeetCode problems solved, reinforcing algorithmic discipline and debugging stamina.
AI systems specialization
Positioned around agents, RAG, MCP tooling, workflow automation, and production GenAI systems.