KHYRON DEVS
All work
Self-initiated technical buildAutomation Systems

JobPilot

An autonomous job intelligence platform featuring multi-provider discovery and deterministic matching.

AUTONOMOUS RECONNAISSANCE & INGESTION RADAR

JOBPILOT TACTICAL FLIGHT DECK

JobPilot autonomous daily opportunity radar
LINKEDIN ATSStripe

Software Engineer

San Francisco, CA
INDEED CRAWLStripe, Inc.

Software Engineer - 2026 Batch

SF / Remote
GREENHOUSE APISTRIPE

SWE I

San Francisco
CANONICAL MASTER LOCKED94.8% PURGED

Stripe • Software Engineer

sha256: 7f3e8b419c2... • SF, CA

3 SOURCES ATTACHED0 REDUNDANT DISPATCHES
JobPilot canonical deduplicated master records table
DEFLECTED • 8 YOEStaff Systems Architect

Requires 8+ YOE, distributed systems leadership.

PURGED FROM FRESHER FEED
CLEARED • 0-1 YOEJunior Frontend Engineer

React, TypeScript, Next.js. Freshers welcome.

COSINE SIMILARITY94.2% MATCH
JobPilot semantic vector cosine distance matching

ZERO-TRUST SSRF FIREWALL

Pre-flight socket IP validation before redirect fetch

DNS RESOLVEDPublic routable IP (104.18.22.42)
PRIVATE CIDR BLOCKED10.0.0.0/8, 172.16.0.0/12, 192.168.0.0/16
CLOUD METADATA DENIED169.254.169.254 AWS/OCI fail-closed
JobPilot production cluster monitoring on Oracle Cloud ARM64
Autonomous Intelligence • Architecture Deep Dive

How JobPilot turns provider spam into an authoritative daily shortlist.

JobPilot eliminates manual job board hunting with an autonomous crawling, deduplicating, and scoring pipeline deployed on Oracle Cloud ARM64. Here is the operational architecture that powers it.

REGRESSION TESTS590 TestsPytest • Worker Leases
DEDUP EFFICIENCY94.8%SHA-256 Canonical Collapse
FRESHER ACCURACY100%Zero Senior Job Leakage
HARDWARE RUNTIME4 OCPU ARM64Oracle Ampere A1 • Docker
ARCHITECTURAL DECK

Core Pipeline Nodes

NODE 01 // CANONICAL DEDUP

Cross-provider noise collapses into a single canonical opportunity.

Raw listings from LinkedIn, Greenhouse, Lever, and Indeed arrive via parallel workers. JobPilot normalizes company slugs, strips promotional noise from job titles, and computes a SHA-256 canonical hash. Duplicate listings collapse into one master record while preserving provider source lineage.

Deduplication Efficiency94.8% Duplicate Elimination
Hash Strategysha256(company:title:location)
Lineage StoreJSONB Provider Reference Array
SHA-256 Canonical HashString NormalizerCross-Provider Lineage
engine/crawler/dedup.py
# engine/crawler/dedup.py
import hashlib, re

def canonical_hash(company: str, title: str, location_type: str) -> str:
    # 1. Strip company legal entity suffixes
    clean_company = re.sub(r"\b(inc|ltd|pvt|corp|technologies)\b", "", company.lower()).strip()
    # 2. Extract core title, removing seniority noise & batch tags
    clean_title = re.sub(r"\b(2025|2026|batch|immediate|urgent|hiring)\b", "", title.lower())
    clean_title = re.sub(r"[^a-z0-9\s]", " ", clean_title).strip()
    norm_tokens = " ".join(sorted(clean_title.split()))

    raw_signature = f"{clean_company}:{norm_tokens}:{location_type.lower()}"
    return hashlib.sha256(raw_signature.encode()).hexdigest()
interactive flight simulator

Canonical Deduplication Engine

COLLISION DETECTED // 1 CANONICAL RECORD
SOURCE A: LINKEDIN CRAWLER
CANONICAL KEY:google:software engineer intern
SOURCE B: GREENHOUSE ATS
CANONICAL KEY:google:software engineer intern
Stream Ingest Rate180 jobs/min
DEDUPLICATION LINEAGE AUDIT
STREAM VELOCITY10800 / hr
DEDUP EFFICIENCY94.8%
DUPLICATES PURGED4095/day
[00:01:24] INGEST: Received 2 raw payloads from LinkedIn & Greenhouse
[00:01:24] NORMALIZE: Stripped suffixes [Pvt Ltd, Inc, 2026 Batch, Summer]
[00:01:24] HASH_COMPUTE: sha256(google:software engineer intern)
[00:01:24] VERDICT: HASH COLLISION DETECTED -> Collapsed to 1 Master Record with 2 source references.
deterministic matcher

Fresher Eligibility & Match Radar

ELIGIBLE // PASSES FRESHER GATE
Candidate Experience Level1 Years YOE
Vector Intent Cosine Alignment75% match
Stack Keyword Match (Python, FastAPI, Postgres, React)3 / 4 matched
88%MATCH SCORE
FRESHER ELIGIBILITY:PASSED (YOE <= 1.5)
SEMANTIC COSINE SIMILARITY:0.75
DELIVERY QUEUE:QUEUED FOR TODAY FEED
network security gate

Fail-Closed SSRF Firewall Simulator

FIREWALL BLOCKED // ZERO TRUST
BLOCKED: METADATA_IP_PROHIBITED (169.254.0.0/16)

Socket connection evaluation before any HTTP request is dispatched by enrichment workers.

169.254.169.254/32BLOCKED (AWS/OCI IMDS)
10.0.0.0/8, 172.16.0.0/12, 192.168.0.0/16BLOCKED (RFC 1918)
127.0.0.0/8BLOCKED (Loopback)
Public FQDN (DNS A-record check)PERMITTED
OPERATIONAL INTERFACE

Production System Gallery

Today / Live Feed
Click to Inspect
Today / Live Feed
Active Radar Feed

Daily ranked fresher-eligible opportunities with match scores and direct apply links.

GET /api/v1/jobs/today/
Canonical Job Archive
Click to Inspect
Canonical Job Archive
PostgreSQL 16 Table

Full deduplicated archive with provider lineage, technology tags, and compensation filters.

GET /api/v1/jobs/archive/
Opportunity Inspection
Click to Inspect
Opportunity Inspection
Deterministic View

Granular eligibility audit, original source provider links, and required skill vector match breakdown.

GET /api/v1/jobs/:slug/
Operator Saved Shortlist
Click to Inspect
Operator Saved Shortlist
User Queue

Locally persisted and server-synchronized bookmark queue for target companies.

GET /api/v1/jobs/saved/
Application Pipeline Tracker
Click to Inspect
Application Pipeline Tracker
State Tracker

Status tracker tracking opportunities from Discovered to Applied to Interviewing.

GET /api/v1/jobs/applied/