Retrieval & knowledge
Grounded answers over documents and structured data—with ingestion, hybrid search, citations, permissions, and freshness designed into the experience.
AI product designer & developer · Texas, USA
Prasiddha K.AI systems builder.
I design AI products that retrieve, reason, act—and show their work.
From retrieval-powered products and agentic workflows to multimodal tools, evals, safeguards, and the interface in between.
Live AI system / HeatSignal
Search anywhere.
Read the risk now.
A working environmental-intelligence demo: resolve any searchable location, pull current public weather and modeled air-quality data, calculate heat risk, and expose every step of the workflow.
Latest available data workflowGlobal search · Public data
01Locate
02Weather
03Air quality
04Classify
05Cite
HEATSIGNAL / READY
Search cities and postal codes worldwide to turn latest-available weather and modeled air-quality data into a transparent heat-risk snapshot.
What I build
AI systems,
end to end.
A useful AI product is more than a model call. It needs reliable context, appropriate tools, clear permissions, measurable quality, and a thoughtful role for human judgment.
Agentic workflows
Tool-using agents that plan, act, retry, and hand off with explicit state, scoped permissions, approval gates, and reliable fallbacks.
Multimodal AI
Interfaces that combine text, images, audio, and structured inputs so people can work naturally across messy real-world information.
Evals & safety
Task-based eval sets, tracing, failure analysis, guardrails, and review queues that reveal when a system is uncertain or needs a person.
Independent product lab
Selected AI work
Four working public-data products that turn live evidence into useful decisions—without hiding the source, freshness, limitations, or role of human judgment.
Live public dataIndependent exploration
/01Voltline
Latest Texas grid intelligence from federal hourly data
Voltline turns EIA-930 hourly submissions for Texas into a clear view of demand, day-ahead forecast, net generation, total interchange, and recent load movement.
01Query EIA-930
02Normalize signals
03Align forecast
04Explain variance
Latest public data · U.S. Energy Information Administration
EIA-930 hourly reportingTexas grid signals monitor
01Query EIA-930
02Normalize signals
03Align forecast
04Explain variance
Connecting to the latest available EIA-930 data for Texas…
How the AI works
A cached read-only service retrieves the public EIA Hourly Electric Grid Monitor feed for ERCO, validates the payload, aligns demand with forecast and other signals by reporting hour, and applies transparent variance and recent-range logic.
Human control & safety
The interface never calls a public-data pattern a reliability alert, reserve margin, or dispatch recommendation. Data latency and official operator ownership stay visible.
How I would evaluate it
Source freshness · schema validation · timestamp integrity · graceful recovery
Live public dataIndependent exploration
/02HeatSignal
Live environmental intelligence for heat-aware decisions anywhere
HeatSignal turns a city, region, or postal code into a current, source-backed snapshot of temperature, heat index, air quality, and practical next steps.
01Resolve the place
02Fetch weather + AQI
03Classify heat risk
04Cite + guide
How the AI works
A client-side orchestration flow geocodes the search, retrieves current weather and modeled air-quality data in parallel, calculates the NWS heat index when conditions meet its threshold, classifies risk, and keeps the data sources visible.
Human control & safety
Modeled data is labeled honestly, missing values remain visible, no personal health outcome is predicted, and official NWS, AirNow, and local guidance stays one click away.
How I would evaluate it
API integrity · location accuracy · calculation tests · failure recovery
Live public dataIndependent exploration
/03FieldLens
A public-data vehicle safety and recall research assistant
FieldLens turns a 17-character VIN into a readable vehicle identity, matching federal recall campaigns, urgent Park It or Park Outside flags, and safe next steps.
01Validate the VIN
02Decode vehicle
03Retrieve campaigns
04Build safety brief
Live public data · National Highway Traffic Safety Administration
No VIN is stored hereVIN safety and recall research assistant
01Validate VIN
02Decode vehicle
03Retrieve campaigns
04Build safety brief
Enter a VIN to turn NHTSA identity data and matching recall campaigns into a readable safety brief.
How the AI works
A two-stage retrieval workflow validates and decodes the VIN through NHTSA vPIC, uses the decoded year, make, and model to retrieve federal recall campaigns, then structures official issue, consequence, and remedy text into a prioritized brief.
Human control & safety
A year/make/model match is never presented as VIN-specific open-recall status. The product sends the VIN only to NHTSA, stores nothing locally, and routes final confirmation to NHTSA or a dealer.
How I would evaluate it
VIN validation · campaign recall · warning fidelity · empty/error states
Live public dataIndependent exploration
/04SceneCraft
A live technical preflight and website risk-signal auditor
SceneCraft inspects any public website and separates technical readiness, accountability hints, external reference signals, RDAP context, header posture, and common scam-risk patterns into explainable results.
01Fetch public page
02Check domain + risk
03Score transparency
04Prioritize next checks
Live public website inspection · no login required
HTML + RDAP + malware DNS + headersWebsite readiness, trust and risk audit
01Fetch public page
02Check domain + risk
03Score transparency
04Prioritize next checks
Enter any public website to inspect technical readiness, accountability hints, external reference signals, RDAP context, header posture, and common scam-risk patterns.
How the AI works
A protected server workflow verifies public DNS on every hop, streams bounded public HTML, follows validated redirects, discovers authoritative registration data through IANA RDAP, compares Cloudflare malware-filter DNS, adds MDN header evidence, and applies versioned deterministic scoring with visible evidence for every warning.
Human control & safety
Positive transparency never cancels a high-impact warning. Results are presented as observable signals—not a truth score, scam probability, safety guarantee, security certification, or accusation about the site operator.
How I would evaluate it
False-positive rate · coverage honesty · signal precision · explanation usefulness
Every product above retrieves latest-available public data and shows its method. HeatSignal uses Open-Meteo and NWS heat-index logic; Voltline uses EIA-930; FieldLens uses NHTSA; SceneCraft inspects a submitted public page, authoritative RDAP registration data, Cloudflare malware-filter DNS, and Mozilla Observatory results. Data can be delayed, revised, incomplete, or unavailable. These are independent products—not client engagements, emergency guidance, professional advice, safety guarantees, or production performance claims.
How I work
Design the loop,
not just the prompt.
The strongest AI experiences make context, state, uncertainty, permissions, and human judgment visible—not magical.
Frame
Define the user decision, source of truth, acceptable failure, and what should remain human.
Ground
Connect trusted context, tools, and permissions. Make citations and provenance part of the product.
Prototype
Build the smallest end-to-end workflow and make the system’s state visible in the interface.
Evaluate
Test quality, cost, latency, safety, and failure modes against representative tasks and a non-AI baseline.
Improve
Add feedback loops, human escalation, monitoring, and versioned changes before expanding autonomy.
Start with the problem
Building
with AI?
If you’re exploring a knowledge product, an agentic workflow, a multimodal experience, or a role where this thinking matters, I’d like to hear what needs to work—and what cannot fail.
Available for focused product design, rapid prototyping, and AI system collaboration from Texas, USA.
Send a project inquiry ↗