Tilak Sadhukhan — Kryvasis

identity :: Founder & CEO ▪ operator :: agentic systems • llm engineering • intelligent product engineering

Agentic AI Engineer specializing in Agentic AI, LLM engineering, AI product architecture, and intelligent software systems. Founder & CEO at Kryvasis, leading product strategy and the development of the next generation of AI-native security system through Generative AI, Agentic AI, and intelligent product engineering. Passionate about building autonomous systems powered by modern AI infrastructure, advanced LLM applications, and multi-agent workflows.

Education

Bachelor of Technology, Computer Science & Engineering

RCC Institute of Information Technology (RCCIIT), Kolkata · 2024–2028

Higher Secondary in Computer

Adamas International School, Kolkata · 2010–2023

Experience

Agentic AI Systems Generative AI Engineering LangChain & LangGraph Retrieval-Augmented Generation (RAG) Prompt Engineering LLM Evaluation AI Product Architecture FastAPI Backend Engineering

Founder & AI Systems Engineer

Kryvasis · 2026–Present

Founded Kryvasis to build the next generation of AI-native security systems powered by Generative AI, Agentic AI, and modern AI infrastructure. Leading the company's product vision, AI strategy, and engineering efforts to develop autonomous security platforms that combine Large Language Models, multi-agent architectures, retrieval-augmented generation (RAG), structured memory, and intelligent tool orchestration. Responsible for driving end-to-end product development—from research, system architecture, and rapid prototyping to production deployment, continuous evaluation, and iterative product improvement—transforming advanced AI capabilities into practical security solutions for modern organizations.

Generative AI Product Engineer

IgniteX Labs · 2026–Present

Researching, prototyping, and shipping AI-native software focused on autonomous reasoning, agent orchestration, multimodal intelligence, and human-AI collaboration. Exploring how LLMs can move beyond conversation into reliable execution through memory, planning, and structured workflows.

Team Lead

GUIDIFY · Dec 2025–Present

Leading an AI-driven career mentorship platform focused on personalized learning and career guidance. Building agentic workflows using LLMs, LangChain, and LangGraph; developing personalized roadmap generation based on user goals, skills, and learning history; integrating conversational AI, skill-gap analysis, and adaptive recommendation systems. Driving product planning, UI/UX decisions, and technical architecture.

Projects

GUIDIFY — AI Career Mentorship Platform

Architected a LangGraph agentic pipeline with specialised agents for skill-gap analysis, roadmap generation, progress tracking, and adaptive recommendation, coordinated via state graphs. Led team to 2nd Runners-Up at Smart India Hackathon 2025 Grand Finale; built React 19 frontend, FastAPI backend, and Supabase PostgreSQL DB with streaming responses and multi-turn session memory.

LangGraph, LangChain, Gemini API, FastAPI, React 19, Supabase, PostgreSQL

VeriFasal — AI Crop Insurance Corroboration System

Multimodal AI + RAG system using Gemma 4 multimodal classification and NDVI satellite imagery to detect fraudulent crop insurance claims via LLM-based corroboration pipelines with structured output generation. Deployed frontend on Vercel and FastAPI backend on Render; completed the full Kaggle hackathon dev lifecycle.

Gemma 4, Multimodal Classification, NDVI, RAG, FastAPI, Python, Vercel, Render

NAT — APK Predictive Examination & Intelligence

6-agent LangGraph pipeline for AI-powered fraudulent Android APK detection: static analysis, permission mapping, behavioural pattern extraction, and LLM-based risk verdict generation. Integrated Androguard deep static analysis feeding structured evidence into a Claude Sonnet LLM for chain-of-thought threat assessment. Designed for GCP Cloud Run deployment.

LangGraph, LangChain, Claude Sonnet, Androguard, React, FastAPI, GCP Cloud Run, Python

ThesisOS — Autonomous Research Workflow OS

Conceptualised a Neo4j knowledge graph over personal PDF libraries for semantic retrieval, citation mapping, and concept relationship visualisation. Designed a LangGraph-based literature review agent with 6 specialised sub-agents: ingestion, entity extraction, contradiction detection, gap analysis, synthesis, and structured report generation.

Neo4j, LangGraph, RAG, Vector Embeddings, PDF processing, Python, FastAPI

Certifications

Oracle Cloud Infrastructure 2025 AI Foundations Associate (1Z0-1122-25) — Score: 98%
Generative AI using LangChain
Data Science with Machine Learning

Achievements

Smart India Hackathon 2025 — 2nd Runners-Up, Grand Finale: Top 3 nationally among thousands of submissions at SKIT Jaipur, Ministry of Skill Development & Entrepreneurship, for the GUIDIFY AI mentorship platform.
Build with Gemma · Kaggle (2026): Completed the full hackathon cycle including deployment and public writeup on Gemma 4 multimodal agricultural AI.
Mind's Eye Pitchathon — Winner (Mar 2025): College-level startup pitching competition.

Skills

AI / Machine Learning

Large Language Models RAG Prompt Engineering Agentic Workflows Multimodal AI Computer Vision Vector Databases Fine-tuning Model Evaluation

Frameworks & Languages

FastAPI React Next.js Node.js LangChain LangGraph Gemini API Hugging Face Transformers Python JavaScript Java C

Data, Cloud & DevOps

PostgreSQL MySQL MongoDB Supabase Vercel Render Oracle Cloud Infrastructure GCP Cloud Run PyTorch NumPy Scikit-Learn YOLOv8

Contact

Email Phone LinkedIn GitHub

Book a Consultation View GitHub