- 🎯 What I do: Build grounded LLM systems: hybrid-search RAG, LangGraph multi-agent pipelines, MCP servers, with Spring Boot / FastAPI backends behind them
- 💼 Experience: Software Engineer Intern @ Lifetrenz · Gen AI Intern @ Dr. Reddy's Foundation · Frontend Dev on Infosys.com @ NeoSOFT (2,200+ pages, WCAG 2.1 AA)
- 🏆 Proof: Shipped at both rounds of the Amadeus × Contentstack "Beyond GenAI" hackathon · Neo4j Certified Professional · 4 Anthropic certifications · Open Source Connect India 2026
- 🧪 Engineering bar: 237 tests on the rate-limiter · retrieval evals in Context Layer · LangSmith tracing · human-in-the-loop AI QA
- 📍 Where: Bengaluru · open to Gen AI / AI Engineer roles · LinkedIn · Email
💻 whoami --verbose (the long version)
vinay@bengaluru:~$ whoami --verbosename: Vinay Kumar Reddy Molakathala
role: Gen AI Engineer · Software Engineer Intern @ Lifetrenz
location: Bengaluru, India 🇮🇳
mission: "Make LLMs grounded, measurable, and actually useful in production."
currently:
building: Context Layer # persistent, model-agnostic memory for AI coding agents (MCP + SurrealDB)
exploring: Graphify # feeding agents a code graph instead of raw code, to beat "lost in the middle"
learning: inference economics, agent evals, GraphRAG
journey:
- 2026.09 → now : Software Engineer Intern @ Lifetrenz
- 2026.04 → 2026.08: Gen AI Intern @ Dr. Reddy's Foundation
- 2025.07 → 2025.12: Frontend Developer (Infosys.com) @ NeoSOFT # 2,200+ pages · WCAG 2.1 AA · AI Text-to-Speech
- 2025.03 → 2025.06: Java Full-Stack Trainee @ EduBridge
- 2023 → 2024 : Python Full-Stack (Techno-hub) · Data Analysis (APSSDC)
badges_earned:
- 🏆 Built at BOTH rounds of the Amadeus × Contentstack "Beyond GenAI" hackathon
- 🧠 Neo4j Certified Professional (cleared at 3 AM, taught Cypher at the workshop that same day)
- 🌱 Open Source Connect India 2026 contributor
- 🎓 Anthropic: MCP · Agent Skills · Claude Code 101 · Claude 101
fun_fact: "My reaction latency at meetup quizzes is higher than the inference latency we spent the day discussing."
Every project below is some remix of this loop: retrieve → reason → cite → measure.
"AI work loses momentum when the conversation gets too long, the model changes, or you switch platforms."
Persistent, model-agnostic project memory for AI coding agents. Structured handoffs in SurrealDB, 17 MCP tools, full-text + vector + hybrid search, token-budgeted context assembly, handoff lineage and a retrieval-evaluation harness.
An assistant that only answers from what it retrieved, and shows you the exact chunks.
Grounded AI chat with citations, keyword + pgvector semantic search, Tika ingestion, Admin/Editor/Viewer RBAC and an audit log. 100% local: Ollama embeddings and answers, so no data leaves the machine.
First 429 at request #101 for a limit of 100, enforced across two separate JVMs.
Six atomic Redis algorithms (fixed window, sliding window, sliding-window counter, token bucket, leaky bucket, concurrency leases). Dynamic admin policies, fail-open/fail-closed modes and 237 green tests.
✈️ AI Travel Assistant · 🏆 Hackathon Part 2
From "plan me a trip" to a budget-aware, day-by-day itinerary in seconds.
A deterministic LangGraph state graph: an intent router fans out to flight, hotel, weather and activity agents, then a composer fans in. Live Duffel + OpenTripMap + OpenWeather data, dual-key Groq failover, and every node traced in LangSmith.
🔎 Enterprise RAG Assistant · 🏆 Hackathon Part 1
Semantic or keyword? Both, fused and tunable live.
A custom Reciprocal Rank Fusion engine over ChromaDB (dense) and BM25 (sparse), with layout-aware Docling ingestion, real-time weight sliders, an RRF score inspector and source-cited answers.
LLMs draft the test cases. Humans approve them. Nothing runs unreviewed.
Generates positive, negative, validation, boundary and business-rule cases, then moves them through Draft → Approve → Execute. Groq, Gemini and Ollama providers with automatic fallback.
Drop in URLs. Ask questions across all of them like they're one document.
n8n webhook ingestion feeds chunking and local Ollama embeddings into a citation-backed RAG on FastAPI, with React and Streamlit front ends.
🧰 Open the toolbox: 13 more builds
- yt-summarizer: MCP server: YouTube transcripts as tools, resources and prompts
FastMCP - Multi-Agent Research System: Search → read → write → critique, self-reviewing reports
LangChain, Groq - n8n Workflows: Zero-paid-API Maps lead scraper, Google-tools AI agent
n8n, Make, Groq - Doc-Intelligent: Classify → extract → summarise, with human verification
Spring Boot, Ollama - Chat with Files: Talk to PDFs, DOCX and XLSX with local models
Spring Boot, Angular - LangChain4j RAG Document Ingestion: Extracts 8 document formats and splits them into RAG-ready chunks, as a CLI or library
Java, LangChain4j - AI Resume Analyzer: ATS-style resume vs. JD scoring in ~2 s
FastAPI, Groq - Hybrid Search RAG: The original RRF hybrid-search prototype
FastAPI, Chroma - TeamTask Manager · 🟢 live: Team task management
Spring Boot, React - E-Commerce Website: 30+ endpoints, JWT RBAC, Razorpay payments
Spring Boot, React - Insight Admin Panel: Full-stack admin dashboard
Spring Boot, Angular - ZymSync: Gym management system
Spring Boot, Angular - SympAI · 🟢 live: Symptom-based disease prediction (85% accuracy)
Flask, scikit-learn - AI Travel Planner: 🐣 Where it all started: my first Gen AI project
FastAPI, Groq
- 🟢 linux-hero ·
JavaScript· pushed 3 days ago
— - 🟣 portfolio ·
TypeScript· pushed 4 days ago
— - 🟣 voice-assistant ·
Python· pushed 12 days ago
Offline real-time dictation for the Linux desktop, powered by Whisper - 🟣 mimu-birthday- ·
HTML· pushed 14 days ago
— - 🟣 Context_Layer ·
HTML· pushed 17 days ago
— - 🟣 quickbite-coo-dashboard ·
Python· pushed 17 days ago
Streamlit + Snowflake Cortex AI dashboard: root-cause analysis of QuickBite's Sep-Oct 2026 refund spike
🏆 Built & volunteered
- Beyond GenAI hackathon Part 1: shipped the Enterprise RAG Assistant
- Beyond GenAI hackathon Part 2: shipped the Travel Assistant, then helped run the event
- Neo4j Agentic AI Workshop: volunteer mentor for Cypher and GraphRAG
- Open Source Connect India 2026: contributor
🎤 Rooms I keep showing up in
AI Dev Circle KSUG.AI null | OWASP Claude Community BLR Snowflake World Tour IBM Bob-a-thon n8n Builder Sessions Local AI & Infra Meetup Self-Learning Agents Mixer Bengaluru Tech Week 2026
✍️ I write up what I learn on LinkedIn: inference engineering, agent memory, self-hosting SLMs, MCP server design, and honest reviews of AI coding tools.
📦 53 original repos · ⭐ 3 stars · ✍️ 39 LinkedIn posts · 🧠 most-used: HTML · JavaScript · Java · Python · TypeScript · 🤖 auto-updated 19 Nov 2026
🔓 Open to Gen AI / AI Engineer roles and collabs on RAG, agents and MCP.
If you're building something where LLMs need to be right, not just fluent, let's talk.
