Skip to content
View Alby-Rajesh's full-sized avatar
1 follower · 4 following

Block or report Alby-Rajesh

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Alby-Rajesh/README.md

Alby Rajesh

AI/ML Engineer  ·  LLM Systems  ·  Knowledge Graphs  ·  Retrieval

Turning unstructured data into queryable knowledge
Agentic Text-to-SQL  •  Text-to-Cypher  •  Graph RAG  •  Schema-constrained LLM pipelines


LinkedIn Email Kerala, India Open to opportunities


About

I'm an AI/ML Engineer at Feathersoft Info Solutions, building production GenAI systems for enterprise clients in agriculture and healthcare. My work sits where language models meet structured data: agents that write SQL and Cypher, pipelines that extract knowledge graphs from documents, and retrieval systems that ground every answer in its source.

Focus

  • LLM agents and tool calling
  • Knowledge graphs and Graph RAG
  • Hybrid retrieval (BM25 + dense)
  • Structured, schema-constrained extraction

Principles

  • Every answer traceable to a source
  • Schemas over prompt hacks
  • Small, tested, observable pipelines
  • Ship it, measure it, improve it

Featured projects

text2cypher-agent

Agentic Text-to-Cypher over Neo4j. A QLoRA fine-tuned Qwen2.5-Coder model writes Cypher; a LangGraph agent validates it with EXPLAIN, self-corrects on errors and answers from the results. Redis cache and memory, LangSmith tracing and execution-accuracy evals.

LangGraph QLoRA Neo4j Redis

voice-text2sql

A voice data analyst for Postgres. Ask out loud, hear the answer, see the SQL, table and chart. LangGraph agent with DSPy SQL generation tuned by MIPROv2 on an execution-accuracy metric; sqlglot AST guard plus read-only transactions.

DSPy LangGraph PostgreSQL Voice

unstructured-2-kg

PDF → ontology-linked Neo4j knowledge graph. Schema-constrained extraction on vLLM, grounded facts, tiered entity resolution and NCBI Taxonomy / ChEBI / LOINC linking. Dockerised, tested in CI.

Python Neo4j vLLM Docling

repo-copilot

An agent for GitHub repositories. Searches code, reads files and commit history, answers with path:line citations, and drafts issues for human approval.

Live demo →

Next.js Groq Supabase

multi-agent-research

Cited research reports from a document collection. LlamaIndex hybrid RAG on pgvector (vector + full-text, RRF) feeding CrewAI crews in three patterns: sequential, manager-led hierarchical, and a self-reviewing reflection Flow. Retrieval and LLM-judge evals.

CrewAI LlamaIndex pgvector

hybrid-rag

Document Q&A with keyword and vector search fused by Reciprocal Rank Fusion, cited answers, and a UI that explains why each passage was retrieved.

Next.js Supabase Gemini

Experience

AI/ML Engineer · Feathersoft Info Solutions  ·  Jul 2025 – present

  • Agentic Text-to-SQL assistant and voice agent for an agriculture enterprise, letting business users query operational data in plain language, by chat or by voice.
  • Patient knowledge graph for a rheumatology platform, with Text-to-Cypher and Graph RAG over clinical data in Neo4j.
  • Autonomous PDF-to-knowledge-graph agent that extracts entities and relationships from documents and loads them into a graph database.

Tech stack

Tech stack

Neo4j vLLM LangChain Hugging Face pgvector Pydantic Docling LangGraph LangSmith DSPy CrewAI LlamaIndex QLoRA fine-tuning

Credentials

  • Claude Certified Developer – Foundations · Anthropic · Sep 2026 · score 910 / 1000
  • B.Tech, Electronics and Communication Engineering · Rajagiri School of Engineering, Kochi · 2021 – 2025

Open to AI/ML, GenAI and backend engineering roles, in India or abroad.
albyrajeshwork@gmail.com  ·  LinkedIn

Pinned Loading

  1. hybrid-rag hybrid-rag Public

    Document Q&A with hybrid retrieval: Postgres full-text + pgvector fused by Reciprocal Rank Fusion, cited answers, explainable evidence.

    TypeScript

  2. repo-copilot repo-copilot Public

    Tool-calling agent for GitHub repos: searches code, reads files and commit history, answers with path:line citations and drafts issues for review.

    TypeScript

  3. unstructured-2-kg unstructured-2-kg Public

    Turn PDFs into an ontology-linked Neo4j knowledge graph: schema-constrained vLLM extraction, grounded facts, hybrid retrieval and entity resolution.

    Python

  4. text2cypher-agent text2cypher-agent Public

    Agentic Text-to-Cypher over Neo4j: QLoRA fine-tuned model, LangGraph validate/self-correct loop, Redis cache + memory, LangSmith evals.

    Python

  5. voice-text2sql voice-text2sql Public

    Voice data analyst: ask Postgres questions out loud. LangGraph agent, DSPy MIPROv2-optimized Text-to-SQL, sqlglot + read-only guards, STT/TTS, auto charts.

    Python

  6. multi-agent-research multi-agent-research Public

    Cited research reports from your documents: LlamaIndex hybrid RAG on pgvector + CrewAI sequential, hierarchical and self-reviewing reflection crews, with retrieval and LLM-judge evals.

    Python