> resuming session

> 4 systems operational

> last deployment: 2026.06.12

> channel open

Advay Sinha

systems engineer · ai infrastructure

DEPTH.01 · SYS.CORE / 01

Neural Core

OPERATORsystems engineer · ai infrastructureoperational

Designs backend systems where behavior under pressure matters more than the demo: multi-agent research pipelines, market simulation, conversational data systems. The recurring problem: how intelligence behaves under constraints.

cap.01
multi-agent orchestration
cap.02
retrieval & evidence grounding
cap.03
backend architecture
cap.04
llm integration & fallback design
cap.05
data pipelines & visualization
cap.06
auth, sessions & persistence

local focus: agent pipelines · evidence-grounded retrievalchannel open

DEPTH.02 · SYS.VAULT / 02

System Vault

4 systems

SYS.MARP / 01evidence-grounded research over a five-stage agent pipelineoperational

Multi-Agentic Research Platform

Answers research questions only with claims it can verify — every stage traced, every citation grounded.

PLANRETRIEVEWRITECRITICVERIFYPGVECTORINOUT

infra:python · postgres + pgvector · gemini embeddings · typescript · docker

> open dossier
SYS.ATS / 02trains and runs trading strategies against live markets, on simulated capitaloperational

Algo Trade Simulator

Lets strategies face real market data without letting simulated capital pretend to be real.

FEEDQUOTESSTRATEGYSIGNALLEDGERCOPILOTINREC

infra:fastapi · react + vite + typescript · mongodb (motor) · yahoo finance api · openai api

> open dossier
SYS.FLOAT / 03conversational explorer for the ARGO ocean-float archiveoperational

FloatChat

Makes a heavyweight scientific archive answerable in plain language, from a cached store that fits on one machine.

NETCDFCONVERTPARQUETQUERYCHATFALLBACKINOUT

infra:python · streamlit · plotly · leaflet · cesium · netcdf → parquet · ollama (optional)

> open dossier
SYS.MOCK / 04AI mock-interview and assessment platform with voice interactionoperational

Mock AI

Generates and grades interviews with an LLM while keeping auth, sessions, and scoring accountable to a real backend.

CLIENTFASTAPIGEMINIEVALUATEMONGOVOICEINREC

infra:fastapi · react 19 · mongodb · gemini 2.0 flash · web speech api · jwt auth

> open dossier

DEPTH.03 · SYS.LOGS / 03

Mission Logs

6 operations · 3 failure traces

  1. SYS.ATS / 02[migration]info

    Session persistence moved to MongoDB

    Accounts, sessions, and simulations migrated from process memory to MongoDB through Motor (async). Process restarts were erasing simulated books mid-run.

    > outcome: sessions survive restarts; the in-memory store retained as a zero-setup path for local runs.

  2. SYS.ATS / 02[failure]warning

    LLM-only command parsing rejected

    Copilot routed simulation commands through the model alone. Loose prompts produced malformed simulations; valid-looking output failed downstream.

    > outcome: hybrid routing shipped — heuristic parsers carry the command path; the model is enrichment.

    experiment closed. the model path is assumed to fail and is routed around, permanently.

  3. SYS.FLOAT / 03[rollback]critical

    Live NetCDF queries rolled back

    Interactive queries against the raw ARGO NetCDF archive exceeded the latency ceiling — multi-second stalls per question made exploration unusable.

    > outcome: rolled back to an offline NetCDF → Parquet converter; the live application reads only the cache.

    conversion is now a precondition, not an optimization. the live path does not exist.

  4. SYS.MARP / 01[architecture]info

    Omnibus prompt decomposed into five agents

    Single-prompt research answers could not be audited — the model answered but could not show its work. Rewritten as Planner / Retriever / Writer / Critic / Verifier with typed trace events at every stage.

    > outcome: each stage inspectable and replaceable; the critique loop bounded by a hard iteration cap.

  5. SYS.MARP / 01[failure]warning

    Retriever executes only the first plan step

    Multi-step retrieval plans truncate to step one; remaining steps are recorded in the trace but never run. Discovered under multi-question research loads.

    > outcome: limit preserved in the execution trace rather than papered over; downstream stages treat thin evidence as a first-class state.

    still unresolved — full plan execution pending.

  6. NEXUS[deployment]info

    Vault pin vs. transformed containing block

    First deployment of the descent engine: the plane transform created a containing block that broke ScrollTrigger's pin fixed-positioning during the Vault traverse.

    > outcome: plane transform cleared for the pinned stratum's focus band; micro-drift sacrificed — the pin is worth more than a 0.5% creep.

DEPTH.04 · SYS.LIVE / 04

Live Systems

telemetry sourcesgithub rest apibuild metadata> aggregated · revalidated 60s

TELEMETRY.DEPLOYbuild lineagedeployed
environment
production
commit
ae5d1b5
branch
master
read at
2026-06-12 · 10:43 utc

> source: vercel build env · stamped at deploy

TELEMETRY.REPOadvay-sinha · last pushesactive
public repositories
16
portfolio
2026-06-12 · 01:07 utc
advay-sinha
2026-06-11 · 23:18 utc
floatchat-ai
2026-05-25 · 17:16 utc
algo-trade-simulator
2026-05-25 · 17:13 utc

> source: github rest api · revalidated 60s

DEPTH.05 · TERMINAL

Terminal Interface

TERMINALmaintenance accessopen
maintenance access layer · direct interface
> type 'help' for the command index

> enter executes · arrow-up recalls · esc releases

DEPTH.06 · NODE: CONTACT

Contact Node

If something here is worth building on, I read every transmission.

CHANNELagent pipelines · evidence-grounded retrievalchannel open
channel
communication channel available
mode
async-first preferred
response latency
variable
domains
orchestration systems · retrieval infrastructure · backend clarity
timezone
asia/kolkata · ist