Algo Theories

Systematic Investment Management  ·  Dubai, UAE

Discipline expressed as code.

Algo Theories is a systematic investment firm. We develop our models in-house, validate them against years of data, and deploy capital against them under a defined risk mandate — our own, and that of a small number of partners.

Request the performance record About the firm
Analysis layers  ·  evaluated in parallel, on every bar Illustrative — not a live signal
0.000 0.236 0.382 0.500 0.618 0.786 1.000 FIBONACCI RANGE HH HL LH LH LL LH EQL BUY-SIDE LIQUIDITY SELL-SIDE LIQUIDITY SWEEP IMBALANCEIMBALANCE BUYER HALT SELLER HALT SELLER HALT PARTICIPATION 926 941 956 970 985 1,000 1,014
01 / 05

Structure

Before anything else, the model establishes whether the market is making progress or standing still — and marks the exact points where that changed.

Define — Measure — Validate — Deploy — Review — Define — Measure — Validate — Deploy — Review — Define — Measure — Validate — Deploy — Review — Define — Measure — Validate — Deploy — Review —
Approach
Systematic
Objective
Absolute return
Discretion
None at execution
Domicile
Dubai, UAE

The firm

Process,
not prediction

We do not take directional views on markets. Capital is committed only when a model the firm has built and validated meets conditions defined in advance, and only within limits set before the position exists.

Every model began as an observation, was specified precisely enough to be tested independently, and was measured across years of data before it was permitted to carry risk. Those that do not survive that process are not deployed.

Our models, and the research behind them, are proprietary and are not described publicly. What we do share with prospective partners, under mutual non-disclosure, is the record of what they have actually done.

Research to allocation

Nothing is deployed
on an opinion

Each stage can return a model to the one before it. Most do not pass the second.

01 / Define

Specify

A hypothesis is written down in terms precise enough that two researchers testing it independently would reach the same result.

02 / Test

Measure

Coded and evaluated across years of history to establish what the model did, rather than what it was hoped to do.

03 / Validate

Prove

Run forward on live data at limited size, with every position and deviation recorded and reviewed against expectation.

04 / Deploy

Allocate

Only then is capital committed, within exposure limits agreed in advance and reviewed on a fixed cycle.

Just Enhance

The firm exists to refine what demonstrably works and to remove everything around it that does not — hesitation, inconsistency, and the judgement calls that quietly cost money.

Capability

Built
in-house

Research, execution, risk and reporting run on infrastructure the firm wrote and maintains itself. Nothing central to the process is licensed in, and nothing is licensed out.

Research

Testing environment

A framework for evaluating candidate models across long histories under consistent assumptions, so results are comparable and reproducible.

Models

Signal generation

Proprietary models run continuously and produce instructions to specification. Their construction is not disclosed.

Execution

Order management

Entry, exit and position sizing are handled to the same specification every time, with no discretion applied at the point of execution.

Risk

Pre-trade controls

Exposure ceilings and per-position limits are enforced by the system itself, before an order can reach the market.

Record

Audit trail

Every event on every account is captured with its full context, producing a complete and independently verifiable record.

In development

Long-horizon model

A research programme aimed at generating candidate models from market data directly, held to the same validation and approval as everything else. Read the direction →

Leadership & governance

A systematic firm is only as sound as the people who decide what gets built, what gets funded, and what gets stopped. Those decisions sit with named individuals here, and they are deliberately not held by the same person.

SHShravan Bachu, Founder and Chief Executive Officer
Shravan Bachu
B.Tech
Founder & Chief Executive Officer

Founded the firm and leads its research. Every model the firm deploys was specified, coded and tested by him before it was handed to anyone else, and he remains the firm's principal quantitative researcher rather than a manager who delegates it.

He sets direction and owns what the firm builds. He does not set the risk limits those models run under.

  • Owns
    Research, model development, firm direction
  • Cannot
    Raise an exposure limit without the Chief Risk Officer
RPRakesh Palakurthi, Chief Risk Officer
Rakesh Palakurthi
M.Sc
Chief Risk Officer & Partner

Holds the firm's risk mandate. Sets exposure ceilings and per-position limits, reviews them on a fixed cycle, and is the only person who can approve a change to them.

Independent of research by design: the person who builds a model is never the person who decides how much capital it may carry.

  • Owns
    Exposure limits, risk review, deployment approval
  • Can
    Halt any model, at any size, without consent
RORohan Revanth, Chief Technology Officer
Rohan Revanth
B.Tech · IIT Guwahati
Chief Technology Officer

Leads the firm's technology. Trained as a computer scientist at IIT Guwahati, with a background in building and evaluating quantitative models, he owns the systems the firm's research runs on — the testing environment, the data pipeline, execution infrastructure and the records behind it.

His mandate is the firm's next generation of research tooling, and he will build and lead the technical team beneath him as that work scales.

  • Owns
    Research infrastructure, data, execution systems
  • Building
    The firm's model-generation programme

Firm structure

Defined before
it is filled

The firm's functions were separated on paper before any of them were staffed. Seats are filled as the book grows — the structure does not change to accommodate whoever happens to join.

Founder & Chief Executive
Shravan Bachu
Appointed

Direct reports to the Chief Executive

Chief Risk Officer
Rakesh Palakurthi · Partner
Appointed
Chief Technology Officer
Rohan Revanth
Appointed
Chief Operating Officer
Defined — open
Head of Trade Surveillance
Defined — open
Finance & Compliance Team
Pradeep Kumar Guntapalli
Appointed
Head of Investor Relations
Reporting to Chief Risk Officer Mohd Inzamam Madhukar Kommineni M Sridhar Naik
Appointed
Technology & Research Engineering
Reporting to Chief Technology Officer
Defined — open
Algo Execution Traders
Reporting to Head of Trade Surveillance Vajja Surendra Adapala Satya Sai Prakash Anumula Raja Ravindra Kumar
Appointed
Appointed Defined, not yet appointed

Separation of duties

What no one
can do alone

Most failures at small systematic firms are not model failures. They are one person holding research, risk and execution at the same time, and quietly moving a limit on a bad day. The structure below exists to make that impossible rather than merely discouraged.

  • Research
    Builds and tests models, and proposes them for deployment. Cannot approve its own deployment.
  • Risk
    Sets and enforces exposure. Approves deployment, and can withdraw it unilaterally at any time.
  • Execution
    Follows the written specification. Holds no discretion to override, resize or skip an instruction.
  • Surveillance
    Reviews executed activity against specification independently of the people who executed it.
  • Investor relations
    Sits beneath risk, not beside it, so what is discussed with an investor is bounded by the mandate rather than by the desire to win the allocation.
  • Reporting
    Drawn from the executing broker rather than from the firm's own systems, so the record is not ours to edit.

Direction

Beyond one
researcher

Everything the firm runs today began with a person noticing something and writing it down. That approach has a ceiling. The programme below is aimed at removing it — and at holding whatever comes out to exactly the same standard as everything that came before.

The problem

Where the
search starts

A researcher can only encode what a researcher has seen. Every model this firm trades exists because someone observed a pattern, wrote it down precisely enough to test, and it survived. That is a sound way to build a book, and it is also a hard limit: the number of hypotheses one person can form in a career is very small next to the number the data contains.

The constraint is not computing power and it is not data availability. It is that the search has always started from human intuition. The programme below changes where the search starts.

The programme

Four stages,
in order

Each stage is a prerequisite for the next. The fourth is the one that matters most, and it is the one most easily skipped.

01 / Foundation

Data at full resolution

Two decades of market history at one-second resolution, aligned with the news and macro record over the same period, held in a form that can be queried consistently rather than sampled by hand.

02 / Generation

Models proposed, not supplied

Rather than testing a rule a researcher already believes in, the system examines the data directly and generates candidate structures of its own — including ones no one thought to look for.

03 / Discipline

Built against overfitting

Anything sufficiently flexible will describe the past perfectly and predict nothing. Candidates are re-run across permutations, periods and market conditions, and are discarded unless they hold outside the data that produced them.

04 / Judgement

The same gate as everything else

A machine-proposed model receives no shortcut. It passes the firm's existing sequence — specify, measure, validate, deploy — and the Chief Risk Officer approves its exposure, or it does not trade.

Boundaries

What this
is not

  • Not autonomous
    No system proposed here allocates capital by itself. Approval remains with the Chief Risk Officer, and that is not a stage we intend to automate.
  • Not prediction
    The objective is a wider search for durable structure, not a claim to know where a market is going. Nothing about the method changes that.
  • Not live
    The programme is in development. It carries no capital today and forms no part of any current mandate.
  • Not a black box
    A model that cannot be stated plainly enough to be tested independently does not pass stage one, regardless of how it was generated.

Forward-looking statement. The programme described on this page is in development and may change materially, be delayed, or be discontinued. It does not form part of any current mandate or offering, and no assurance is given as to its outcome, performance or timing.

Investors

A small book,
run conservatively

Algo Theories deploys principally its own capital. We accept a limited number of partners where the fit is right — investors seeking moderate, evidenced returns from a defined mandate rather than a growth narrative.

  • Mandate
    A defined set of models, documented in advance, with no discretionary deviation once capital is committed.
  • Risk
    Risk per position and maximum exposure are agreed in writing before deployment, and reviewed on a fixed cycle.
  • Reporting
    Statements drawn from the executing broker rather than from our own systems, with the full position log available behind them.
  • Evidence
    The complete executed record is made available on request under mutual non-disclosure. Returns are not published publicly.

Request the record

Introduce yourself and the mandate you are considering. Where there is a fit, we will share the executed record and arrange a call.

contact@algotheories.com