# phavella · Custom AI, built for production.

> phavella is an AI solutions studio. We design, build, and integrate custom AI into the stack a company already runs: agents, apps, voice, document intelligence, analytics, decisioning, and creative pipelines. Use cases big and small, from a single workflow to a full platform. A human signs off on every consequential decision; the routine majority auto-executes. Then we stay until it works in production.

## Positioning

- **What we are:** A senior, hands-on AI solutions studio: four named principals who build and run the systems themselves. No juniors on production systems.
- **What we build:** Custom AI, end to end: full-stack products, multi-agent systems, voice interfaces, retrieval at scale, dashboards, decision systems, and creative pipelines. We build from scratch or integrate into the stack you already run (CRM, inbox, docs, ads, ops).
- **How we work:** Scope, Build, Ship, Operate. Weekly demos on your data, not slides. Evals and monitoring wired in. Routine work auto-executes; decisions that are expensive to get wrong route to a named human reviewer with context and an audit trail.
- **What we won't do:** Let AI make a consequential decision alone. Rip out your system of record. Free pilots. Equity-only deals. Strategy decks without a build behind them. Engagements where the math doesn't work.

## The nine practices

1. **AI Agents & Workflow Automation** · `https://phavella.com/capabilities/agents-automation` · Multi-agent systems and automations that take routine work off your team's plate. 618+ automations shipped across 12 industries; a 34-tool MCP server for Meta Ads.
2. **AI-Powered Apps & Products** · `https://phavella.com/capabilities/ai-apps` · Full-stack products with AI at the core. Our ad-production platform runs a 13-state production pipeline; 15 production AI apps built at Newsweek.
3. **Conversational & Voice AI** · `https://phavella.com/capabilities/conversational-ai` · Assistants, copilots, and voice interfaces. A financial chatbot across 4 agent frameworks at a wealth-management fintech; 20 medical AI agents at a clinical-intelligence company; production voice pipelines in our ad-production platform.
4. **Document Intelligence & Enterprise Search** · `https://phavella.com/capabilities/document-intelligence` · Search, extraction, and understanding across millions of documents. A 30M-document Elasticsearch index at Newsweek; financial OCR for PE at a private-markets data platform; 66GB of retail catalogs at a construction-commerce marketplace; 100k+ clinical codes at a clinical-intelligence company.
5. **Data, Analytics & Dashboards** · `https://phavella.com/capabilities/data-analytics` · Pipelines, executive dashboards, and analytics. Pharma analytics at a life-sciences commercial-analytics firm for three top-10 pharmaceutical companies; portfolio operations for two North American private-equity firms; embedded data work at a Fortune-tier technology company and a major airline.
6. **Risk, Fraud & Decision Systems** · `https://phavella.com/capabilities/decision-systems` · Scoring, underwriting, and decisioning that survives regulators and scale. A $150M risk portfolio and $1B annual transaction policy at Drip Capital; 16 million fraud decisions a month at Citibank, 5+ years in production.
7. **Content & Creative Automation** · `https://phavella.com/capabilities/creative-automation` · Generative pipelines for video, ads, and content at production volume. Our ad-production platform produces video ads end to end across 11 video models at 5× the weekly output; Creative OS shipped a 522-commit client ad platform.
8. **System Integrations & AI Modernization** · `https://phavella.com/capabilities/integrations-modernization` · AI wired into the stack you already run: CRM, inbox, docs, ads, ops. Slack, Google Drive, Sheets, Docs, Dropbox, and Meta Ads integrations shipped inside our ad-production platform; a 34-tool Meta Ads MCP server; embedded modernization inside a Fortune-tier technology company and PE operating companies.
9. **Evaluation & Reliability** · `https://phavella.com/capabilities/evaluation-reliability` · The discipline that makes AI safe to run in production. BFCL, TAU-bench, and DeepEval harnesses; multi-LLM comparison; hallucination reduction; monitoring. Built into everything we ship; available standalone.

## The bench

Four named, senior principals. The edge is the systems; the people are the proof.

- **Parik Ahlawat** · Principal · Agents & Orchestration · `https://phavella.com/bench/parik` · Multi-agent platforms, model routing, n8n orchestration. 618+ automations across 12 industries; creator of Kronus; built an in-house ad-production platform (13-state pipeline, 5× ad output) and Creative OS (522 commits).
- **Arpit Jhamb** · Principal · Data & Portfolio Ops · `https://phavella.com/bench/arpit` · ETL, executive dashboards, and business automation across operating companies. Past: a life-sciences commercial-analytics firm (three top-10 pharmaceutical companies), a North American private-equity firm, a mid-market private-equity firm, a Fortune-tier technology company (government affairs, 1+ yr embedded), a major airline.
- **Nikhil Bery** · Principal · RAG & Document AI · `https://phavella.com/bench/nikhil` · RAG, document intelligence, multi-LLM pipelines, vector + graph search at scale. Past: Newsweek (5 years, 30M-doc index, 15 production apps), a clinical-intelligence company (20 medical agents, 100k+ clinical codes), a wealth-management fintech, CGIAR (9k+ records), a construction-commerce marketplace (66GB), a private-markets data platform (PE financial OCR).
- **Shashwat Verma** · Principal · GenAI & Financial Risk · `https://phavella.com/bench/shashwat` · GenAI for risk and policy design. $150M risk portfolio at Drip Capital, $1B annual transaction policy designed; 16M Citibank fraud decisions/month over 5+ years.

## Reference builds (case studies)

Four systems, told properly: what broke, what we built, what changed. Every number traces to shipped work.

- **Fraud decisioning at scale** · `https://phavella.com/work/citibank-fraud` · Citibank. 16 million fraud decisions a month across major retail co-brand card portfolios; bank-grade model validation (KS, PSI); edge cases route to analysts. 5+ years in production.
- **Risk portfolio & transaction policy** · `https://phavella.com/work/drip-risk` · Drip Capital. A $150M risk portfolio managed and the policy governing $1B in annual transactions designed; exposures above policy limits route to a risk lead before funds move.
- **Document intelligence at newsroom scale** · `https://phavella.com/work/newsweek-doc-ai` · Newsweek. A 30M-document search index behind 15 production AI apps; editors approve before anything publishes; ~497 hours saved every 30 days.
- **AI video ad production pipeline** · `https://phavella.com/work/ad-production-platform` · Ad Platform (in-house). A 13-state pipeline that scripts, generates across 11 video models, and voices ads end to end, with human approval gates before anything reaches Meta Ads; 5× weekly ad output.

Also shipped by our principals: Kronus, Creative OS, a clinical-intelligence company, a wealth-management fintech, CGIAR, a construction-commerce marketplace, a private-markets data platform, a life-sciences commercial-analytics firm, a major airline, a North American private-equity firm, a mid-market private-equity firm, a Fortune-tier technology company.

## How we engage

Scope. Build. Ship. Operate. A named process with a human on the calls that matter. Details: `https://phavella.com/engagement`.

1. **Scope**: One call, then a written plan: the use case, the data, the integration points, the number it has to move.
2. **Build**: Principals build. Weekly demos on your data, not slides. An eval suite comes with it.
3. **Ship**: Into your stack, your infra, your access controls, with evals and monitoring wired in.
4. **Operate**: We run it with you until it's boring. Then we hand it over, or keep operating it.

Pricing and billing shape are conversation-only, never on public surfaces.

## Thinking (essays)

Notes from building AI that has to work. `https://phavella.com/thinking`

- **Integrate, don't replace** · `https://phavella.com/thinking/integrate-dont-replace` · Layer AI on the stack you already run, don't migrate first. The ad platform added 34 operations over the existing API, zero migrations.
- **What 1,322 automations taught us about 30 industries** · `https://phavella.com/thinking/what-1322-automations-taught-us` · 1,322 automations across 30 industries: the shape repeats for the easy 70%. The last mile never generalises, and that is where projects die.
- **Bank-grade is a discipline, not a badge** · `https://phavella.com/thinking/bank-grade-is-a-discipline` · 16M fraud decisions a month teaches what bank-grade means: validation, stability monitoring, fairness review, and a written policy layer.
- **Retrieval at 30 million documents** · `https://phavella.com/thinking/retrieval-at-30-million-documents` · 30M documents, 15 production apps: search quality is an evaluation problem before an infrastructure one. It recovered ~497 hours in 30 days.
- **In regulated care, design the escalation first** · `https://phavella.com/thinking/healthcare-ai-escalation-first` · Twenty medical agents in production: design the escalation first, ground them in 100k+ clinical codes, and staff a handoff a human answers.
- **Volume is a system property** · `https://phavella.com/thinking/volume-is-a-system-property` · 5× the ad output came from a 13-state pipeline and 11 routed models, not better prompts. You cannot parallelise what you have not named.
- **Why most AI never reaches production** · `https://phavella.com/thinking/why-most-ai-never-ships` · The gap between a demo and a system that runs, and what closes it.
- **Human-in-the-loop is a feature, not a disclaimer** · `https://phavella.com/thinking/human-in-the-loop-is-a-feature` · Where to put the human, not whether to have one.
- **Strategy that ships** · `https://phavella.com/thinking/strategy-that-ships` · Discovery that writes code beats a deck that doesn't get acted on.
- **What AI can actually automate in a 3PL back office (and what it can't)** · `https://phavella.com/thinking/automating-the-3pl-back-office` · The 3PL back office is a rekeying machine. AI takes the reading, entering, and flagging; the rate exceptions and disputes stay human, layered over the TMS.
- **Document intelligence for clinic groups: intake without the retyping** · `https://phavella.com/thinking/document-intelligence-for-clinics` · Clinics type patient data three times. The model reads the referral and drafts the PMS entry; staff approve, nothing patient-facing runs unattended.
- **The AI app you ship in eight weeks: what production-ready means** · `https://phavella.com/thinking/what-production-ready-means` · Production-ready is not a demo: the eval set first, monitoring, error paths, a human review lane, the boring 90% handled. Scope, Build, Ship, Operate.
- **From dashboards to decisions: what AI analytics is for** · `https://phavella.com/thinking/from-dashboards-to-decisions` · A dashboard shows the past and waits. AI analytics proposes the decision and its reason. Drip: a policy governing $1B that holds per invoice.
- **Fraud screening at 16 million decisions a month: what holds** · `https://phavella.com/thinking/fraud-screening-what-holds` · 16M fraud decisions a month for 5+ years: what holds is never the model. It is the thresholds, the review lane, the audit trail, the drift evals.
- **Why your e-commerce ad pipeline stalls at ten variations** · `https://phavella.com/thinking/why-ad-pipelines-stall` · Hand-run creative stalls around ten a week. A 13-state pipeline across 11 models lifted output 5x. The bottleneck is production, not ideas.
- **Private AI, in-house: the shift to internal models and agentic infrastructure** · `https://phavella.com/thinking/private-ai-in-house` · Regulated finance, healthcare, and government are moving AI in-house: open models, agentic infrastructure, and a human on the consequential call.
- **Document intake in an accounting practice: what actually eats the hours** · `https://phavella.com/thinking/document-intake-for-accounting-practices` · Client records arrive in every format a client feels like sending, and somebody re-keys them. A system can read the pile and file it. A qualified person still signs anything a regulator will read.

## Start

Tell us the use case, big or small, custom or straightforward. A principal reads every brief. `https://phavella.com/start`

## Company

- **Website** · https://phavella.com
- **Contact** · https://phavella.com/start