Work

Everything I have shipped, newest first, including what I shut down. The deeper ones link to a case study.

Aeranow · 2026 to now

An operations brain for multi-unit operators. The current project builds it into an autonomous org for Kalamandir Jewellers, a ~₹5,850 crore jewellery house, starting on the vendor and procurement side.

early deployment on the group's real data
Meridian OSprototype · Sep 2026

An autonomous organization for a door, frame and hardware distributor. Humans write the Charter; agents hold seats and spend within limits.

82 tests, 6 graded trials, 3 red-team trapsgithub.com/parth012001/meridian-osread the case study →
Voice carrier deskprivate · Aug 2026

Inbound carrier sales as a production voice deployment against a legacy TMS that speaks raw TCP. The agent is never told the rate ceiling.

43 real calls, 9 bookings, 0 without an identity checkread the case study →
Memory for an AI companionprivate · Sep 2026

A memory architecture for a companion that has to remember you between conversations. Threads, not facts, as the unit; a stable belief block the reply path never recomputes; validity windows for facts that stop being true.

17-page design, 4 figures, a survey of six memory systemsread the case study →
Carrier Deskshipped · Aug 2026

An AI carrier sales rep that negotiates inside a ceiling it cannot see. The eval caught the model arguing itself into double-brokering; the tool layer stopped it.

610 offline tests, 6 adversarial personas, 28 logged decisionsgithub.com/parth012001/carrier-deskread the case study →
Toglo, LMS deliveryprivate · May to Jul 2026

First discovery conversation with an HVAC franchise to a live LTI 1.3 integration inside their production Moodle in under 90 days.

go-live 21 Jul 2026, 300+ tests, 19-check live harnessread the case study →
Toglo, FieldMateprivate · Apr to Jun 2026

Voice and camera coaching agent for HVAC technicians. A deterministic gate that a too-agreeable model cannot open.

intent benchmark 35/45 to 44/45, 34-phrase bilingual adversarial corpusread the case study →
Greenlightshipped · Jul 2026

An IT agent that acts when policy allows, asks a human when it doesn't, and earns autonomy per exact action shape, then loses it on one bad run.

hash-chained audit log, 50-action replay before any rule widensgithub.com/parth012001/greenlightread the case study →
TaskFlowshipped · Jun 2026

Internal task, KPI and daily-planning platform with four-pillar productivity scoring.

Captain AIsunset · Jun 2025 to Apr 2026

A chief of staff that sends the email and books the meeting instead of drafting, behind an approval gate on anything risky. Paid for by account executives for pre-call briefs.

140 beta users, 90% of proposed actions acceptedgithub.com/parth012001/capAIread the case study →
Atlas AIshipped · Sep 2025

AI travel concierge.

E2B MCP servermerged upstream · Aug 2025

Sandboxed Python and JavaScript execution, files and packages as 7 MCP tools. Merged into Klavis (YC X25).

CodeLabshipped · Aug 2025

Coding-education platform exposed as 13 MCP tools behind an AI gateway so agents can grade, hint and run interviews.

Curishipped · Jul 2025

Conversational product research over reviews and specs.

Kalyxasunset · Aug 2024 to Jun 2025

AI styling marketplace, web and iOS, Stripe Connect payouts. Wound down June 2025. It taught me to pick a wedge and de-risk the hard part before building.

20+ releases, 75% user growth, $150K seed commitment
Boeing UAVshipped · Aug to Dec 2023

Real-time object detection for autonomous UAVs on Jetson edge devices, plus ROS2 navigation and obstacle avoidance, with an 8-person team.

95% detection at 30+ FPS, 85% lower simulated collision riskgithub.com/parth012001/BoeingUAVPhase2