# Sachit Bansal: full profile > Sachit Bansal is an AI engineer from IIT Mandi who builds agentic AI systems, intelligent automations and AI-powered applications, and takes them from concept to deployment. 2× KrackHack champion; administers 50+ websites for 3,000+ users. Available for freelance projects, internships and Werkstudent roles. Website: https://sachitbansal.sbssdigital.com Role: AI Engineer · Agentic AI · Intelligent Automation · Software Engineering ## Why work with Sachit: Not just writing code. Building solutions. I believe good engineering starts with understanding the problem, not choosing a technology. 1. **Understand the problem**: I look beyond the initial requirement to the underlying challenge, the people using the solution and the outcome that matters. 2. **Engineer the solution**: I bring together AI, software development, APIs, data pipelines and automation to design solutions that are practical and scalable. 3. **Build, deploy and improve**: I own implementation, integration, testing and deployment, so it works beyond a demo. The objective isn't to use AI everywhere. It's to use technology where it creates measurable value. ## Services ### Agentic AI & intelligent assistants AI systems that go beyond answering questions. - Autonomous AI agents and multi-agent workflows - Document intelligence and knowledge assistants - Retrieval-Augmented Generation (RAG) - LLM tool calling and intelligent information retrieval - Research and decision-support systems - Example: An AI research assistant that retrieves information, verifies claims and provides evidence-backed answers. ### Business process automation Reduce repetitive work and create intelligent workflows. - AI-powered workflow automation - API integrations and backend automation - Intelligent document and data extraction - Automated reporting and information processing - Real-time data processing and decision systems - Example: Automating a multi-step business workflow that currently needs repetitive manual work. ### AI-powered applications & products Turn ideas into working digital products. - AI-powered web applications - Custom dashboards and internal tools - Backend and API development - Mobile applications - AI integrations into existing products - Example: A complete application with AI capabilities, backend architecture and production deployment. ### AI infrastructure, deployment & engineering Build systems that can operate reliably. - Python backend development - Docker-based deployments - Cloud infrastructure and hosting - Linux administration and server security - API integrations and production support - Example: Deploying and maintaining applications with the right infrastructure and security practices. ## Engineering case studies ### SciAgent: An agentic AI research assistant that verifies what it says. Agentic AI · Research, 2026 - Challenge: AI assistants can produce convincing answers without checking whether the underlying information supports their claims. In research, that is a critical limitation. - What was engineered: An agentic research assistant that searches academic papers, retrieves relevant evidence, evaluates claims and gives grounded responses. - Five specialised LLM tools - BGE embeddings and FAISS retrieval - Cross-encoder reranking - Natural Language Inference (NLI) for claim verification - A framework-free LLM tool-calling loop - Validation: 100% citation accuracy on a hand-annotated benchmark; 100% abstention accuracy; 0.92 held-out fact coverage - What it demonstrates: Engineering AI that goes beyond generating text: retrieval, reasoning, verification and reliability. - Technology: Python · Agentic AI · FAISS · BGE · Cross-encoders · NLI - Source: https://github.com/Sachitbansal/researchAgent ### Customer Targeting: Real-time customer intelligence through streaming data and machine learning. Streaming ML · Inter IIT Tech Meet 14.0, 2025 - Challenge: Traditional customer segmentation goes stale as behaviour changes. Businesses need to understand customers as transactions happen. - What was engineered: A distributed streaming pipeline that processes transaction events and maintains a live Customer 360 state. - Stateful joins and reducers - Incremental feature updates - Distributed streaming architecture - Online machine-learning inference - Concept-drift handling - Validation: ≈0.83 streaming accuracy - What it demonstrates: Distributed systems, streaming data, machine learning and real-time decision intelligence. - Technology: Pathway · Python · Streaming systems · Online ML - Source: https://github.com/Sachitbansal/CustomerTargetting ### PodClip: Turning long-form podcasts into short-form content intelligence. Applied ML · Speech, 2025 - Challenge: Finding the engaging moments in long podcast recordings takes hours of manual listening and editing. - What was engineered: A backend pipeline that ingests YouTube podcasts, extracts audio, transcribes and separates speakers, and identifies high-engagement moments. - Audio extraction and processing - Whisper-based transcription - Speaker diarization - Sentiment analysis - Vector indexing and content intelligence - What it demonstrates: Combining applied ML, speech processing and backend engineering into a useful content workflow. - Technology: Python · Whisper · Hugging Face · Applied ML - Source: https://github.com/Sachitbansal/PodClip ### MyItihas: Production mobile app development, from engineering to app-store release. Mobile · Production, Dec 2025 – Feb 2026 - Challenge: A social mobile app needs more than screens: authentication, data models, APIs, user interactions and reliable deployment. - What was engineered: Owned the full product development lifecycle and independently shipped production builds to Google Play and the Apple App Store. - Peer-to-peer and group chat - Authentication and profile management - Follower / following system - Backend data modelling - API integration for search and chat storage - What it demonstrates: End-to-end product ownership, mobile engineering, backend integration and production execution. - Technology: Flutter · Supabase ### Webspark: An AI-powered website builder. AI agents · Hackathon, 2025 - Challenge: Getting a website made is slow and expensive for a small business. - What was engineered: A system that turns a plain-language instruction into a working website through AI-assisted design, development and deployment. - Natural-language brief to site plan - AI agents for design and code - Automated deployment - What it demonstrates: Orchestrating multiple AI agents into one product flow, built and demoed under hackathon pressure. - Technology: Python · AI agents · Next.js ## Journey - **Class 4, Rubik's cubes**: A simple cube turned into a fascination with harder and harder puzzles, and the satisfaction of solving them. - **Age 14, Curiosity turns to code**: Started learning Python, web development and mobile app development, not for a curriculum but to make ideas real. - **Age 16, First app on the Play Store**: A full year building and publishing an Android app alone, every line by hand, before AI coding tools existed. - **Class 10, Keynote at GDG, UPES Dehradun**: Delivered a hands-on Flutter session as a keynote speaker at a Google Developer Group event. - **Lockdown, Competitions & hackathons**: Started competing online, including events run by IIT Bombay, IIT Guwahati and IIT Madras. - **Lockdown, 2,000+ learners on YouTube**: Started a channel teaching Python and Flutter that grew organically past 2,000 subscribers. - **IIT Mandi, From curiosity to engineering**: Cybersecurity, AI, machine learning and systems. Google Cybersecurity certificate, Cybersecurity Club, server administration, mentoring and teaching. - **Today, Agentic systems & automation**: Software engineering, applied AI, agentic systems and automation, with the same motivation: understand deeply, build something useful. ## Education - Indian Institute of Technology, Mandi: B.Tech, General Engineering, specialisation in AI & Robotics, 2024 – 2028 - Leibniz University Hannover, Germany: Exchange semester, Computer Science & AI, Oct 2026 – Mar 2027 ## Achievements - 1st place, KrackHack 3.0 (GDG IIT Mandi · 2026) - 1st place, KrackHack 2.0 (GDG IIT Mandi · 2025) - 1st place, Space Code 2026 (IIT Mandi · 2026) - AIR 9 among 15,000+ (Shaastra Juniors, IIT Madras) - 1st place, Case Study Competition (Shaastra Juniors, IIT Madras) - Google Cybersecurity Professional Certificate (Google) ## Leadership and contribution - **System Administrator, SAIC, IIT Mandi**: Administer 50+ websites on the primary student server for 3,000+ students and faculty: Docker deployments, DNS and server security. - **Tutor, PRAYAS 3.0**: Mentored 100+ students in hands-on Arduino and machine learning. - **Core Team, Kamand Prompt, Programming Club**: Delivered a web development session to 500+ students. ## How a project works 1. **Understand**: We discuss the business problem, the intended users and what success should look like. 2. **Architect**: I pick the right technology, design the solution and define a practical development plan. 3. **Build**: I develop in focused iterations, validate functionality and share progress. 4. **Deploy**: I integrate, test and deploy the solution into its intended environment. 5. **Improve**: I help refine the product based on feedback, performance and actual usage. ## Contact - Email: sachitbansaldeveloper@gmail.com - LinkedIn: https://linkedin.com/in/sachit-bansal/ - GitHub: https://github.com/Sachitbansal - X: https://x.com/sachitbansal20 - Based in: Hannover, Germany (available for remote collaboration internationally)