v0 · IN ACTIVE DEVELOPMENT

Placement automation, built on AI from the ground up

DriveStream is a modular platform automating campus recruitment — from CV compliance to drive management — targeting a 95%+ cut in manual vetting turnaround time.

95%+

Projected reduction in
vetting turnaround time

See what's in development Get in touch

vetting_pipeline.ts live schema

Extract

claims

Flag

guidelines

Verify

proofs

Review

human

Why we built this

Built by people who ran the drives

DriveStream is built by a team of placement coordinators from SRCC's (Shri Ram College of Commerce) placement cell. We've spent recruitment seasons inside the actual bottlenecks: auditing hundreds of resumes line-by-line by hand, chasing a backlog of grievance emails with no central system, and working with legacy placement software too rigid to configure for a single drive's specific rules.

That experience shaped DriveStream's core design principle: placement automation should adapt to how a cell actually works, not force a cell to adapt to the software.

Placement Cell Experience

2,000+ Student Drives Observed

Built From the Inside

The Vision

One core engine, configured your way

DriveStream isn't a fixed placement tool — it's a configurable framework. The rules, workflows, and policies of any placement cell can be built on top of the same underlying engine.

Configurable Policies

Set eligibility rules, CGPA cutoffs, and course restrictions per drive, not per platform-wide default.

Adaptable Workflows

Vetting logic, grievance routing, and drive stages configure to how your cell actually operates.

Built to Extend

New modules plug into the same core engine as placement needs evolve — not a rebuild each time.

Current Status

In Development

Here's what's actually built so far

DriveStream's core framework is in active development. Rather than a polished demo, here's real architecture: a configurable placement framework built from 12 core components, forming the foundation every future module will plug into.

956+

Lines of code

Core components

Stages in the vetting pipeline

Platform (Together AI) powering inference + fine-tuning

Built

Vetting & Compliance

Claim extraction, guideline flagging, proof verification, human review.

Built

Feedback & Learning

Correction logging, few-shot injection, fine-tuning pipeline.

Built

Core Infrastructure

PDF extraction, shared LLM client, benchmarking.

Planned

Drive & Admin Tooling

Drive lifecycle management, coordinator dashboards, and grievance routing — next up on the roadmap.

Deep Dive

Inside the vetting engine

The most fully-built part of DriveStream today — a 4-stage pipeline that reads a campus's actual compliance guidelines and applies them consistently, claim by claim.

STAGE 01

Claim Extraction

Every checkable statement in a CV gets pulled out individually — a metric, an award, a role, a certification — rather than treating the resume as one undifferentiated block of text.

STAGE 02

Guideline Flagging

Each claim is checked against the campus's actual compliance rules to decide: does this need supporting proof, or does it already break a formatting/wording rule outright?

STAGE 03

Proof Verification

When a student submits evidence — an offer letter, a certificate, a transcript — this stage checks whether it actually substantiates the specific claim: matching numbers, dates, and titles, not just confirming a document exists.

STAGE 04

Human Review

Anything ambiguous or low-confidence routes to a placement coordinator for the final call. The system handles the volume; a person handles the judgment calls.

Feedback loop

Every correction a coordinator makes gets logged with the reasoning behind it, and automatically improves the system's next decision — with a clear path to full model fine-tuning as corrections accumulate.

Built on Llama models served via Together AI.

Interested in DriveStream?

Whether you're curious about the architecture, want to follow progress, or have placement bottlenecks of your own — get in touch.