workfOS is the operating system for your AI workforce — specialist AI team members that run on your schedule, connect to your tools, and handle the knowledge work your team shouldn't have to. Today: inbox monitoring, candidate screening, document research. And that's just the beginning.
You have the answer somewhere — in a policy doc, a past proposal, a report from six months ago. Finding it takes longer than it should. Research Assistant reads your documents so you don't have to search.
You posted a role. 200 résumés arrived. The strongest candidate is in there — but they didn't keyword-stuff their way to the top. The Recruiter scores for substance, not SEO.
A good opportunity landed in your inbox last Tuesday. You didn't see it until Friday. By then, someone else had already responded. Job Scout reads every message so you don't miss the ones that matter.
You can tell when an email was written by AI. So can the person receiving yours. Communications Manager drafts in your voice — trained on your writing, not the internet's average.
WHO IT'S FOR
PROPERTY MANAGEMENT
Maintenance requests, lease inquiries, vendor follow-ups — all arriving in the same place. An AI worker monitors your inbox and surfaces what needs your attention. Research Assistant answers tenant questions from your policy documents. Your mornings stop starting with triage.
BOUTIQUE CONSULTING
Every hour spent searching internal documents, prepping briefing materials, or screening subcontractor résumés is an hour not billed. Research Assistant works through your document library. The Recruiter filters your candidate inbox. Your team focuses on the work clients pay for.
JOB SEEKER
You're looking for the right one. Job Scout monitors every inbox, reads every recruiter email, and surfaces only the ones worth your time. Twice a day, a digest. No noise.
Choose a specialized AI worker from the roster
Connect your data sources and models in minutes
Your worker runs on schedule, 24/7
THE ARCHITECTURE
workfOS matches each job to the right model tier — frontier APIs, open-source, fine-tuned, or ensemble consensus.
Claude, GPT-4o — best-in-class for general reasoning and language tasks
Local open-source via Ollama (Llama, Mistral, Qwen). HuggingFace domain specialists — roadmap.
LoRA-adapted models trained on your data — voice, domain, workflow
Same prompt → multiple models → synthesized best response — a platform-wide routing capability. Live in Research Assistant; extends platform-wide.
At full maturity, the agency selects the tier at worker configuration time — based on job complexity, data availability, and cost envelope. Today, frontier and local models route through a unified BYOK router; ensemble synthesis is live in Research Assistant and extends platform-wide.
Built today
YOUR WORKFORCE
Each specialist is a different routing decision — different data, different signals, different model tier.
Answers questions about your documents — grounded, cited, no hallucinated sources.
Design Decisions
Pipeline
01 · INGEST
PDF / URL / Markdown
Chunked & embedded on ingest (nomic-embed-text)
02 · STORE
ChromaDB
Per-worker isolated store, under customer data root
03 · RETRIEVE
Similarity + rerank
Threshold-gated; low-relevance chunks dropped
04 · ANSWER
Cited synthesis
Frontier or local LLM, multi-model ensemble
Answers grounded in your documents, with citations.
BUILT BY
Anupam Chandra (AC)
Chief Architect · Founder, SoftExcel Technologies
32-year career designing systems at scale — GSA ($400M eCommerce modernization), Fannie Mae (2.5M loans/hour processing), Capital One (founding Object Factory). Recent work: AI/ML entity resolution across 80M+ product records, enterprise search platforms, and now workfOS.
Available for strategic contract engagements.
12+
users in private alpha
18
workers running
9
mailboxes connected
PRICING
Flat monthly rate. Your API keys. No surprises. BYOK
Free during alpha
Pricing at launch
Pricing at launch
Contact us