AI Engineer · Hyderabad, India
I engineeragentic systems
Open-source contributor with published libraries and models, focused on production reliability, observability, and cost-efficient systems.
the canvas listens to your cursor01Who I am
About
I turn frontier AI into systems enterprises can trust in production.
Software Engineer with expertise in AI, experienced in building and shipping enterprise-grade LLM applications — agentic systems, multi-agent orchestration, RAG pipelines, and knowledge-graph reasoning. I combine strong backend engineering with hands-on GenAI development, from LLM fine-tuning and prompt engineering to scalable APIs and production deployment.
Open-source contributor with published libraries and models, focused on production reliability, observability, and cost-efficient systems.
Agentic AIMulti-Agent OrchestrationRAG PipelinesKnowledge GraphsLLM Fine-TuningBackend Engineering
At a glance
- Role
- AI Engineer
- Company
- Tata Consultancy Services
- Based in
- Hyderabad, India
- Education
- B.Tech, Computer Science and Engineering
02What I work with
Skills & Tools
Languages
04PythonJavaCSQL
GenAI & Agentic AI
09LangChainLangGraphMCPA2ALLMsLLM Fine-TuningAgentic WorkflowsPrompt EngineeringTool Calling
RAG, Search & Graphs
08RAG PipelinesFAISSChroma DBEmbeddingsSemantic SearchNeo4jCypherElasticsearch
Reliability & Observability
05GuardrailsLangfuseTracingToken & Cost TrackingHuman-in-the-Loop
Backend & Tools
09FastAPISQLAlchemyAlembicJWT / OAuth2DockerGitAWSMongoDBPostgres
03Where I've worked
Experience
TCS
AI Engineer
Tata Consultancy Services
Feb 2025 — PresentHyderabad, India
Agentic Decision-Intelligence Platform for Semiconductor Supply Chain
PythonPandasLangGraphNeo4jRAGLiteLLMNVIDIA NeMoFastAPIDockerLangfuse
- Delivered an agentic AI platform that acts as a command center for supply-chain executives — live dashboards, a conversational insights chatbot, proactive recommendations, and automated workflows — shortening the path from disruption to decision.
- Modeled the end-to-end supply chain — suppliers, fabs, assembly and test, logistics, and customers — as a dynamic Neo4j knowledge graph, combining multi-hop Cypher reasoning with RAG so agents surface risks, bottlenecks, and single-source dependencies in real time.
- Implemented Intelligent Choice Architecture (ICA): agents frame vetted decision options with trade-offs and trigger automated workflows, so executives act on recommendations rather than raw data.
- Hardened the platform for enterprise use with guardrails on agent outputs and Langfuse observability for end-to-end tracing, debugging, and per-query token and cost tracking.
Multi-Agent Orchestration Framework
LangGraphMCPAzure OpenAIPythonFastAPI
- Engineered a framework-agnostic orchestration system where independent agents, tools, and workflows register dynamically and compose into pipelines — via MCP and custom communication protocols.
- Implemented human-in-the-loop approval gates for high-impact agent actions, making autonomous decisions auditable and safe for enterprise AI workflows.
Codebase Documentation RAG Chatbot
RAGFAISSFastAPIPythonAzure OpenAI
- Built a RAG assistant that generates, indexes, and retrieves technical documentation directly from source-code repositories — owning ingestion, chunking, embeddings, and FAISS-backed semantic search end to end.
- Enabled natural-language querying of APIs, architecture, and code behavior, accelerating developer onboarding and cutting time spent navigating unfamiliar codebases.
04Things I've built
Selected Projects
Featured
cyphersmith
Open-source text-to-Cypher Python library
PythonLiteLLMNeo4jCyVerLoRA
- Authored and published a pip-installable library that converts natural-language questions into validated, read-only Cypher and executes them on Neo4j, with LiteLLM-based provider-agnostic support for all LLM providers and local models.
- Engineered a layered safety pipeline — read-only keyword/procedure blocking plus CyVer syntax, schema, and property validation with retries — and shipped an interactive CLI with schema-aware prompting, injectable business context, and live progress logs.
- Fine-tuned and published a Qwen3.5-4B LoRA adapter for Neo4j text-to-Cypher generation on Hugging Face, complementing the library with an open model.
Content Publishing Platform
Fully async REST backend for a blog platform
FastAPIAsync SQLAlchemyPostgreSQLAlembicJWT / OAuth2PydanticAWS S3
- Built a fully async REST backend for a blog platform — users, posts, profile images, and password resets — with FastAPI and SQLAlchemy 2.x async ORM, using Pydantic validation and Alembic-managed schema migrations across PostgreSQL and SQLite.
- Implemented end-to-end security: Argon2 password hashing, OAuth2/JWT bearer authentication, and owner-only authorization on protected resources, plus background SMTP password-reset emails and Pillow image processing with S3-backed storage.
Multi-Agent Log Analysis System
AI-powered log intelligence platform
Google VertexRAGFAISSFastAPIElasticsearch
- Developed an AI-powered log intelligence platform where specialized agents collaborate on log parsing, anomaly detection, failure analysis, and root-cause investigation.
- Enabled natural-language investigation of system behavior, error patterns, and incidents through RAG over Elasticsearch-backed retrieval, reducing manual log triage effort.
05Foundations
Education & Certifications
Education
Saintgits College of Engineering
B.Tech in Computer Science and Engineering
2020 — 2024Kottayam, Kerala
Certifications
AWS Certified Cloud Practitioner
Amazon Web Services
06Get in touch
Let's make something resonate.
I'm open to AI Engineer roles, agentic-AI collaborations, and interesting open-source work. The fastest way to reach me is email.
aswin.kv202@gmail.com