TRUEFOUNDRY × POLARIS HACKATHON

Build Resilient Agents. Ship ones that survive production.

A one-day, in-person hackathon in Bangalore. Round 1 is open to everyone online. Shortlisted builders get a seat at the Polaris campus on 26 September to build against real failure modes: flaky tools, bad evals, runaway cost, no observability.

SAT 26 SEP 2026 BANGALORE · POLARIS CAMPUS IN-PERSON · ONE DAY TWO ROUNDS PRIZE POOL: TBA
Register for Round 1 See what you'll build
event.status
// WHAT IT IS

One day, one room, 350–400 builders

TrueFoundry and Polaris are co-hosting a day-long hackathon in Bangalore. It follows the public launch of TrueFoundry's agent harness — this is the developer moment that comes after it.

Entry runs in two rounds. Round 1 is an open online form: tell us why you want in and what you plan to build. We shortlist from those answers. Round 2 is the build itself, in person, on 26 September, using TrueFoundry's agent harness and AI gateway alongside whatever else your stack needs.

No prior TrueFoundry experience is required. Bring an idea and a laptop.

DATE
Sat, 26 Sept 2026
VENUE
Polaris campus, Bangalore
FORMAT
In-person, one day
SCALE
350–400 builders
ENTRY
Free · shortlisted
PRIZE POOL
₹ TBA
// THEME · WORKING TITLE

Resilient Agents

Agents that hold up in production. THEME BEING FINALISED

Demo agents work on the happy path. Production agents meet timeouts, rate limits, malformed tool output, ambiguous user intent, and a cost ceiling. Most break there, and most teams find out from a user rather than a dashboard.

That gap is the theme. Resilience is an engineering property, not a prompt: retries and fallbacks that respect idempotency, traces you can replay, evals that catch a regression before a release, guardrails at the boundary, and a latency and cost budget the system actually honours under load.

TrueFoundry's agent harness and AI gateway exist for exactly this: the harness runs and instruments agents, the gateway sits in front of models to route, cap, and observe traffic. Both are available to you on the day.

// WHAT YOU'LL BUILD

Six directions, all fair game

These are starting points, not categories you must pick from. Anything that makes an agent more dependable counts.

01 · FAILURE RECOVERY
Agents that recover from tool failures
A tool 500s, times out, or returns malformed JSON. Build retry and fallback logic that degrades gracefully instead of hallucinating a result — with replay so a half-finished run resumes rather than restarts.
02 · OBSERVABILITY
Traces you can actually debug
Span-level visibility into an agent run: which tool was called, with what arguments, how many tokens it burned, where it stalled. Make a multi-step failure readable in under a minute.
03 · EVALS
Evals that catch regressions
Build a harness that scores agent trajectories, not just final answers. Ground truth sets, LLM-as-judge with agreement checks, and a gate that blocks a bad change in CI.
04 · GUARDRAILS
Guardrails at the boundary
Prompt injection through retrieved content, tool calls with unsafe arguments, PII leaking into logs. Enforce policy on input and output, and make a blocked action explainable.
05 · COST & LATENCY
Budgets that hold under load
Model routing by task difficulty, caching, speculative or parallel tool calls, hard token ceilings per run. Show p95 latency and cost per task before and after.
06 · ORCHESTRATION
Multi-agent systems that don't deadlock
Handoffs between specialised agents with clear contracts, shared state that stays consistent, and supervision that kills a runaway loop instead of paying for it.
// HOW IT WORKS

Two rounds

Round 1 is online and open to all. Round 2 is in person, by invitation.

01
Round 1 — register online
A short form: who you are, why you want to join, and what you plan to build. No prototype needed at this stage. Registration and submissions run on a hackathon platform — PLATFORM TBA
02
Post about it (optional, worth doing)
Share what you want to build on LinkedIn or X and tag us. Public posts carry extra incentives in review — details confirmed with the Round 1 form.
03
Shortlist
We read every entry and shortlist on the strength and specificity of the build idea, not credentials. Seats are capped at 350–400. Shortlisted builders get an email invitation with venue details.
04
Round 2 — build in person
Saturday 26 September at the Polaris campus in Bangalore. Build on the day, demo to judges, ship something that runs. Food, power, and wifi are on us.
05
Demo and judging
Live demos at the end of the day. Working code beats slides. Judging criteria published with the Round 1 form.
// DAY-OF AGENDA

Saturday, 26 September

Shape of the day. All times are provisional and marked TBC until the final schedule is published.

MORNING · TBC
Check-in and setupBadges, wifi, workspace. Get your environment running before the clock starts.
MORNING · TBC
Kickoff and problem briefThe theme in detail, judging criteria, and what "resilient" means for scoring.
MORNING · TBC
Technical walkthrough: agent harness + AI gatewayHow to instrument, route, and evaluate agents with the TrueFoundry stack. Credentials handed out here.
MIDDAY · TBC
Build starts · lunch servedMentors from TrueFoundry and Polaris on the floor for the rest of the day.
AFTERNOON · TBC
Mentor checkpointOptional review of your approach and failure-mode coverage.
EVENING · TBC
Submissions closeCode and a short write-up on the hackathon platform.
EVENING · TBC
Demos and judgingShortlisted teams demo live to the judging panel.
EVENING · TBC
Results and prizesWinners announced on stage.
// PRIZES

Cash prize pool

There is a real cash pool. The figure is being finalised and will be published before Round 1 closes.

FIRST PLACE
₹ TBA
Cash prize · amount to be announced
SECOND PLACE
₹ TBA
Cash prize · amount to be announced
THIRD PLACE
₹ TBA
Cash prize · amount to be announced

Additional category awards and partner credits are being confirmed. Prize pool to be announced.