Skip to content

24-hour offline hackathon · Mangaluru, Karnataka · Oct 30–31, 2026

TechSiege

24 hours. 200+ builders. One campus. Ship real agentic AI — not chatbot wrappers.

--
Days
--
Hrs
--
Min
--
Sec

Not another chatbot hackathon

TechSiege rewards systems that perceive, plan, call tools, remember, and finish real work — evaluated against an explicit agentic bar.

0+
Participants
Engineering & MCA students
0+
Teams
2–4 members per team
0
Tracks
Real agentic problem statements
0 hrs
Offline Build
Mangaluru campus venue

Prompt wrappers don't pass

A single LLM call behind a chat UI — no tools, no memory, no multi-step behaviour — gets screened out, no matter how polished it looks.

Agents that act do

Show a system that reads the world, plans across steps, remembers context, recovers from errors — and leaves an audit trail. That wins.

Six arenas. Pick your mission.

Every track expects working agentic behaviour — tool use, planning and memory — applied to a real problem.

  1. Autonomous AI

    Agents that plan, act and finish work end-to-end.

    Research agents · Productivity copilots · Enterprise workflow automation

  2. AI for Education

    Tutors and tools that actually teach and administer.

    Personalized tutors · Assessment graders · Teacher admin assistants

  3. AI for Healthcare

    Information & workflow assistants for care teams.

    Triage intake agents · Literature research · Discharge-summary helpers

  4. AI for Finance

    Agents that read numbers and explain what matters.

    Financial research · Expense analysis · Business-intelligence bots

  5. AI for Social Impact

    Accessibility, public-service and planet-first agents.

    Accessibility aides · Civic-service bots · Environmental monitors

  6. AI Developer Agents

    Agents that ship software alongside you.

    Code + test generation · Auto-documentation · DevOps & code review

The bar for evaluation

Screening judges check every submission against this list. Miss the core, miss the finals.

  1. Tool / API integration

    Agent calls real tools, APIs or data sources — not just chat replies.

  2. Multi-step execution

    Planning, task decomposition or workflow orchestration across steps.

  3. Persistent state / memory

    Session state, long-term memory or context carried across runs.

  4. RAG / external knowledge

    Retrieval over docs, web or datasets grounding the agent's actions.

  5. Multi-agent or human-in-the-loop

    Agent collaboration, delegation, approvals or escalation paths.

  6. Safety & auditability

    Error handling, validation, sandboxing and a visible action log.

Timeline

24 hours, minute by minute

08:00 AM kickoff on Day 1 → awards by 05:30 PM on Day 2. Tabs switch between days.

  1. Registration & Check-in

    Team verification, IDs, kits, workspace allotment.

  2. Opening Ceremony

    Welcome, rules, tracks, judging and safety briefing.

  3. Keynote: The Agentic Shift

    From chatbots to tool-using autonomous systems.

  4. Hacking Begins

    Team formation locked · repos created · mentor floor opens.

  5. Lunch + Mentor Checkpoint 1

    Architecture reviews: tools, memory and scope.

  6. Mentor Checkpoint 2

    Mid-build review: working tool call required.

  7. Dinner + Night Sprint

    Fuel up. Quiet hacking hours with mentor support.

  8. Midnight Checkpoint

    Progress log + demo-of-something-working submission.

How winners are picked

Weighted rubric, published upfront. Agentic depth counts the most.

Agentic Capability25%

Tool use, planning, memory, autonomy depth.

Innovation20%

Originality of the problem + approach.

Technical Implementation20%

Architecture, code quality, robustness.

Problem Relevance & Clarity15%

Real need, clear scope, measurable outcome.

User Experience10%

Usability, design polish, accessibility.

Demo & Presentation10%

Live demo quality + storytelling.

A technically brilliant demo with no tool use, memory or planning caps at 75. Build the agent, not the slides.

Awards

Glory and prize pool

Indicative amounts — the final pool scales with sponsorships. Every finalist gets stage time.

Featured

Grand Champion

Best overall agentic system.

₹50,000* + Trophy

  • 1st Runner-Up

    Outstanding execution & demo.

    ₹25,000*
  • 2nd Runner-Up

    Strong agent + story.

    ₹15,000*
  • Best AI Agent

    Deepest autonomy: tools, memory, planning.

    ₹10,000*
  • Most Innovative Solution

    Boldest idea, well executed.

    ₹10,000*
  • Best Education Agent

    Top build on the Education track.

    ₹7,500*
  • Best Social Impact Agent

    Top build for public good.

    ₹7,500*
  • Best Developer Agent

    Top dev-tooling agent.

    ₹7,500*
  • Best Autonomous Workflow

    Cleanest end-to-end orchestration.

    ₹7,500*
  • Best UX

    Most usable, accessible interface.

    ₹5,000*
  • People's Choice

    Audience-voted favourite demo.

    Goodies + Glory

What you ship on Day 2

Submission closes 11:00 AM. Incomplete entries don't advance to screening.

  1. Public GitHub repo with a clear README (setup, env, run steps)

  2. Project summary: problem, users, agent capabilities (≤ 300 words)

  3. Agent architecture diagram (tools, memory, flow, models)

  4. 2–3 min demo video (unlisted link) showing real tool use

  5. Declaration of APIs, models, datasets & pre-existing code used

Portal link shared at kickoff · Questions? cynex.depticb@gmail.com

Sponsors — announcing soon

We're lining up partners who want to meet 200+ builders. Check back shortly.

Sponsor wall coming soon.

Tiers and confirmed logos will appear here once announced.

FAQ

Questions, answered

Students from engineering/MCA colleges across Mangaluru, Udupi/Manipal, Karnataka and Kerala. Teams of 2–4 members. Beginners with strong fundamentals are welcome — mentors will help you level up.

Final call

50+ teams.
One arena. 24 hours.

Registrations close when slots fill. Oct 30–31, 2026 · Alva's Institute of Engineering and Technology, Mijar Campus · Auditorium. Bring your crew, your keys, your ambition.

--
Days
--
Hrs
--
Min
--
Sec