DeepEmergence · Confidential

Building Silicon Mind for Silicon Employees.

Ajay Pratap Singh · Founder, CEO & CTO · ajay@deepemergence.com · deepemergence.com

DeepEmergence × BoldCap1

Market gap · cognition vs wrappers

LLM alone ≠ mind.

Domain AI apps sell copilots, agents, and “workers” by wrapping an LLM with tools and skills. Demos look employed. Companies still hire humans.

LLMs + tools != an autonomous worker. The winner of the AI race will be whoever successfully engineers miniaturized human cognitive functions into a single, scalable digital employee.

Mind parts LLM alone misses
Language is one slice of cognition
  • Language → LLM — fluent; not sufficient for labour
  • Vision / depth / fusion — seeing + sensing still open
  • Imagination → world models — simulate before acting
  • Memory ≠ chat history — must change the base of self
  • LLMs are stateless — context is attached, not lived
  • No self / world awareness — weak internal + external state
  • Causal reasoning — cause / effect & interventions missing from next-token prediction
  • Goals & beliefs — don’t evolve or self-correct
  • Skill learning / RL — little real-world action learning
  • Broader brain stack — ML/NN parts AGI needs are absent
What’s sold today
OpenClaw · Hermes · Paperclip · domain agents
  • Wrapper pattern — LLM + tools + skills for a domain
  • Prompt / cron bound — reactive or scheduled
  • Loop ≠ mind — act cycles, no durable self
  • Loop engineering — scaffolds “long work” that never leaves toy projects
  • No causal reasoning — correlates & continues; doesn’t intervene on cause
  • Happy-path demos — works when the world is framed
  • Human babysitter — judgment, stop, repair
  • No mandate / liability — not a hireable employee
  • Siloed domain bots — finance ≠ ops ≠ support
  • Pilot cliff — demos die in messy orgs
Engineered silicon mind
Compose the missing parts → hireable labour
  • Composed cognition — not chat seats with APIs
  • Causal reasoning — intervene on cause, not just continue the pattern
  • Linear memory since first boot — unbroken lived history
  • Always available — working toward goals, not waiting on a prompt
  • Continuous mind — owns the day across signals
  • Durable identity — memory that updates the self
  • Goals & beliefs — evolve; anti-drift under mandate
  • World-model foresight — imagine, then act
  • Remote-employee surface — laptop + phone native
  • Governed long-horizon work — accountable like a hire
  • Miniaturised mind functions — engineer first, then generalise
  • Digital worker bar — where world expectation is heading

Thesis: AGI is an engineering problem.

DeepEmergence · Confidential2

Conviction · signals by mind function · application layer

Labs ship mind ingredients. We ship the digital employee.

Frontier work is capturing parts of an AGI-like mind — beyond LLM agents. Causal reasoning is the most important ingredient — and it already exists as serious research (not vapour). Directional signals only (not partnerships). We are the application layer: compose those ingredients into a hireable digital employee.

Mind functionEarly signals (class of player)Ingredient role
Causal reasoningMicrosoft DoWhy / PyWhy-class — causal discovery, identification, estimation; Pearl-style intervention stacks at big-tech + academiaCause → effect; intervene, not only correlate
LanguageFrontier LLMs — OpenAI / Anthropic / Google-classFluent reasoning & dialogue
Imagination / dynamicsWorld models — DeepMind Genie-class; JEPA-style latent predictionSimulate before acting
Vision–spatialWorld Labs–class spatial intelligence; multimodal perception stacksSee & situate in space
Memory / predictionPersistent-memory & representation research across labs & papersState beyond the prompt
Planning / controlLong-horizon agents in sim; robotics & autonomy programsGoals → actions over time
Computer controlCUA / computer-use agents — Anthropic-class, Operator-classClick, type, navigate UIs
EmbodimentRobotics + sim platforms (e.g. NVIDIA Cosmos–class tooling)Act in the physical world
Our layer

Digital employee — compose ingredients into identity, mandate, lived memory, org graph, and configurable roles. Customizable for enterprise & retail. Engineered path toward AGI as a system, not a single model.

Why we still win

Labs race foundations. We productise the hire into services/headcount budgets. Fuel improves us; they do not own employment. First-wave application layer — not out-training Genie or GPT.

Conviction: Surf the wave of digital-mind foundations. Capture them as product — silicon employees — while the industry is still stuck on copilots.

DeepEmergence · Confidential3

What we engineer · not another domain wrapper

Sophon — the composed mind under every role.

Sophon is one engineered runtime that turns a swappable model into a governed digital employee. The model is a component; the system is the product. Build the mind once — then configure roles on top.

Roles — configured, not rebuilt Digital employees: retail + enterprise

Chronos, Sira, Elyra… and finance / legal / ops. A new role is config on one runtime — not a 12th wrapper to maintain.

Sophon harness · the moat Identity · guardrails · hooks · evals · replayable traces

Every action is scoped, policy-bound, and audited. Deny-by-default; humans approve high-risk writes. This layer is the liability surface.

Beyond-LLM cognition World model · causal engine · typed memory · planner

Simulate consequences before acting, reason cause→effect (not correlation), remember with lineage — what an LLM alone can't do.

Foundation models — the fuel Swappable & commoditized

Any frontier LLM + retrieval, hot-swapped behind evals. We route intelligence; we don't retrain it.

Not a wrapper.

A wrapper is model + prompt + tools + session memory. Sophon simulates and verifies before any irreversible write, and reasons about cause — a system, not a chat loop.

Hard to copy.

Value compounds where models don't deprecate: causal schemas, eval history, trace archive, and the org coordination graph — plus owned identity and the customer data plane.

Labs won't eat it.

Frontier labs race the bottom layer (fuel). We own the top three — the employment surface, governance, and customer data. Every better model only makes Sophon stronger.

LLM is fuel for language. Sophon is the engine that makes labour possible — and the part that survives every model release.

DeepEmergence · Confidential4

Competition · who sells the hire today

We compete with employers of people — not AI copilots.

Copilots fight over the software slice. We compete for the headcount & services slice — the same budget as IT, consultancy, and staffing that supply employees for hire.

Typical enterprise opex mix
~55%
Headcount & servicesPeople, contractors, BPO, consulting
~10%
Software & toolsSaaS, copilots, agents, platforms
~35%
Other opexFacilities, infra, marketing, travel, legal, G&A, COGS-adjacent…

Illustrative — mix varies by industry. We aim at the people slice; copilots fight software; other opex is real but not our category.

IT & BPO / GCC labour

Services & captives that staff knowledge roles at scale — default remote work capacity.

Consultancy & managed services

Humans embedded in workflows — billed as delivery teams, not software seats.

Staffing & talent platforms

Staff aug, PEOs, remote marketplaces — the employment surface we digitise.

Why not AI agent vendors: they compete for the ~10% software aisle. Labour outcomes still buy from the ~55% headcount & services slice — that is our category.

DeepEmergence · Confidential5

Market · Global · remote-capable digital employees

One mind. Configurable roles. A global labour market.

We aim at remote-capable human jobs — almost every knowledge role a laptop & phone can do — sold as customizable digital employees. Figures below are wage-pool TAM, citeable SAM, and company SOM targets.

TAM · Global
Remote-capable wage pool
$8.4T

Worldwide employer cost of knowledge roles that can run on laptop + phone — the ceiling as configurable roles cover nearly all such jobs.

Wage pool · not software
SAM · Global
AI application software
$172B

Gartner worldwide AI application software spend (2025) — the citeable near-term budget aisle we re-price as hireable digital employees.

Longer ceiling · IT services $1.69T
SOM · Global
Yr-5 revenue target
$85M

DeepEmergence Year-5 global revenue target across retail roles and enterprise deployments — company plan, not a third-party forecast.

Plan · not market research
How the funnel works

TAM = remote-capable knowledge wage pool. SAM = Gartner AI application software 2025 ($172B) as citeable near-term budget; longer ceiling is IT services ($1.69T). SOM = DE Year-5 revenue target — not a third-party forecast.

Sources: BEA/FRED private wages · Gartner IT & AI spend 2025 · NASSCOM FY25 · see market-methodology.md. SOM = company target.

DeepEmergence · Confidential6

Phase 1 · products · ~18 - 24 months

What we will have built.

One engineered mind — then an ecosystem of digital employees on top. Same Sophon core for retail and enterprise.

Sophon — core of the digital mind

Identity · continuous cognition · lived memory · causal foresight · org channels. Model-agnostic runtime on laptop + phone — the substrate under every hire.

Silver · Phase 1
  • Chronos — AI Chief of Staff
  • Sira — equity advisor
  • Elyra — healthspan COO
  • Saarthi — career operator
  • Lux — personal finance
  • Butler — home & lifestyle
Gold · preview
  • Auvi — family office
  • Kat — relationships
  • Guru — AI teacher
  • Vael — leadership coach
  • Jolly — legal literacy
  • Butler Gold — human + workforce
Enterprise
  • Digital FTE — scoped hire
  • Team module — 3–12 FTEs / dept
  • Ghost Org — 8 depts · one graph
  • Hybrid org — human + silicon
  • Ledger — finance FTE pack
  • Chronos Team — exec CoS module
  • Studio — custom roles & evals
  • Rampart — auth · MCP · fail-closed

Phase 1 ships Sophon + first retail seats + first enterprise ladder — not twelve separate bots.

DeepEmergence · Confidential7

Phase 1 · $1.5M pre-seed · 18–24 months

Compute > salaries — India-weighted USD. GPU, APIs, experiments, and eng/AI software lead; top talent cash; lean GTM + reserve.

Use of fundsShareWhy this line exists
Compute, cloud & software46% · $695KGPU · frontier APIs · experiments · eng/AI tools — largest
Team salaries37% · $550KTop India AI rates · 5 eng + 5 researchers · founder $6K
Workplace & ops8% · $115KLaptops · office · coffee · insurance · entity
Marketing & sales (GTM)4% · $55KLean proof GTM — paid Chronos + LOI
Pilots & delivery3% · $40KOne hybrid design partner path
Reserve & misc3% · $45KHiring lag · compute spikes · random events
Milestone 1

Continuous-mind Sophon beta — identity, memory, governed action in daily use.

Milestone 2

Paid Chronos users + Sira beta — demand for hireable roles.

Milestone 3

1 enterprise pilot / LOI — hybrid design partner.

Ask: $1.5M pre-seed → Phase 1 proof. Next capital after hireability is real.

DeepEmergence · Confidential8

Progression · engineering the rest of the mind

Depth → twin → body.

After Phase 1 proof (Now → Q4 ’26): fuse more signals, then permissioned judgment, then a physical body. Same engineering thesis — miniaturised mind functions, then generalise. Phase 4 has no date.

Phase 22027
Continuous depth
What it continuously understands — not only what you tell it
  • Personal context mesh — calendar, inbox, docs, finance, health, relationships
  • Enterprise context mesh — CRM, ERP, tickets, policy, team state
  • Continuous work loop — observe → believe → prioritise → act / escalate
  • Expect — works before the prompt; software-first on devices you already own
Phase 32027–28
Digital twin & platform
Delegate low-ROI decisions · keep human moments
  • Personal delegation — career, shopping, travel, household admin shortlists
  • Enterprise autonomy — planning, vendors, obligations — policy-bounded
  • Platform APIs — career, compatibility, workforce primitives for partners
  • Expect — permissioned judgment layer; twin screens the long tail
Phase 4Horizon
Physical body
Real silicon employee — no date
  • Same mind, new surface — sensors, mobility, actuation on Sophon
  • Beyond screens — spaces, logistics, care, field ops in the world
  • Earn embodiment — mind + trust first; hardware does not lead
  • Vision — digital employee → embodied silicon employee

Not a product roadmap of wrappers. A progress curve for the silicon mind: richer state → delegated judgment → embodiment.

DeepEmergence · Confidential9

Team · who engineers the digital mind

Ajay Pratap Singh · Founder, CEO & CTO

IIT Madras · YC · Goldman. Premji Invest · Owns product, platform, GTM, raise. Ships the employment surface; knows wrappers are not enough — Sophon still needs dedicated research ownership.

Co-founder · Chief Research Officer (CRO)

IIT Madras PhD candidates in discussion; IIT / MIT / Stanford preferred.

DeepEmergence · Confidential10

Advisors · four seats · all open

Open seats. Forming the bench.

Four advisor seats — none filled. Seat 1 is in consideration with two research professors; seats 2–4 are open.

#SeatStatusFocus
1AI Research ProfessorOpen · considering 2Prof. Balaraman Ravindran · Prof. Mitesh Khapra — mind stack, RL, memory, planning
2Applied AI Technical LeaderOpenProduction systems — what breaks when loops leave the sandbox
3Product & SalesOpenHireable digital labour GTM — packaging, pilots, close
4Software Companies GrowthOpenScaling software businesses — distribution, expansion, ops leverage

DeepEmergence · Confidential11

Why BoldCap

You back frontier AI. We engineer the mind layer.

BoldCap is an AI-native seed fund for deeply technical Indian founders building toward AGI — at conviction, not consensus. Your market map already names AI teammates. Sophon is that substrate: one composed mind, many hireable roles.

Pramaana Labs · verification

You co-hosted the Formal Verification Summit and back Pramaana. We engineer verify-before-write and causal reasoning into every digital employee — labour that doesn't just say sorry.

Your utopian worlds

Finance AI · Agent Identity · Formal Verification — portfolio lanes we extend. We ship the employment runtime (Sophon + Chronos, Sira, Ghost Org) — not another single-purpose agent.

Fit to BoldCap

Pre-seed/seed · IIT technical founder · AGI as systems engineering · India → global from day one. Your pit crew — follow-on capital (portfolio raised ~$100M+), founding talent, forward-deployed storytelling — matches a category we are defining: silicon employees.

DeepEmergence · Confidential12

DeepEmergence

We are creating digital employees for hire.

LLM alone ≠ mind. Without a mind, you are not hiring an employee — you are building a wrapper.

AGI is an engineering problem. Sophon composes the missing architectural pieces into a unified silicon mind. We scale this mind through a compounding 5-stage roadmap:

  • Phase 0 (Core Engine) — Building Sophon — a vanilla cognitive core engineered to natively control computers and process raw signal I/O.
  • Phase 1 (The Software Org) — Orchestrating autonomous roles, agent teams, and hybrid human–AI workflows.
  • Phase 2 (Hardware Integration) — Embedding this mind into edge hardware to integrate natively into lives and enterprises.
  • Phase 3 (The Behavioral Twin) — Leveraging hyper-dense data collections to build a digital twin that deeply understands and mimics user behavior.
  • Phase 4 (Physical Embodiment) — Porting this exact same, hyper-contextualized mind into a physical robotic body.

Global market expectation is heading here inevitably. Sophon is building the architecture.

deepemergence.com · ajay@deepemergence.com · Ajay Pratap Singh · CEO & CTO

DeepEmergence · Confidential13