Institutional Research
A full institutional research desk across three research divisions, covering equities, commodities, FX, crypto and macro on demand.
Sephiance Qant is an institutional research house built on a proven quantitative engine. Refined through continuous optimisation and tuning, the platform operates across three research divisions. Every number is produced by a deterministic engine grounded in published mathematics. Every input is cross-referenced across multiple sources, independently verified, and overlaid with our own proprietary analysis before it informs a conclusion.
The result combines the breadth and speed of cutting-edge AI with the discipline and reproducibility of institutional quant, delivered on demand.
Sephiance Qant covers five asset classes: equities, fixed income, commodities, currencies, and crypto, across three dedicated research divisions. Each division operates independently, then passes its output to the founder, who sets the house view and makes the final call before anything reaches a client.
Equity and portfolio research for client capital. Builds fundamental coverage on single names, constructs and optimises multi-asset portfolios, and maintains ongoing performance surveillance against mandate benchmarks.
The markets desk covers all five asset classes: equities, commodities, FX, crypto, and macro. Cross-asset surveillance is augmented by a dedicated Greek-market specialist who contributes local on-the-ground intelligence.
Independent quantitative oversight. Every thesis, signal, and recommendation from Divisions I and II is validated here through point-in-time backtesting, factor decomposition, and automated quantitative checks before any output is cleared for release.
Above the three research divisions sits the founder, whose background spans private equity boutiques and family offices, with deep grounding in quantitative finance, portfolio management, and cross-asset markets. Every research output from Divisions I, II, and III (every report, every signal, every recommendation) passes across his desk for final review before a single call is released to a client.
Behind every report is a team of specialised agent-analysts that runs a full desk cycle each trading day. They do not simply summarise the news. They take a position, size it, defend it against an independent risk committee, and only then release a call.
Before the bell, the desk takes in live prices, global news and sentiment across equities, commodities, FX, rates and crypto, and reads the broader macro climate.
Asset & Wealth, Global Banking & Markets and the Greek-market specialist each run their analysts, then argue bull against bear. Disagreement is the point: a thesis that cannot survive the debate never reaches the trader.
The Trade Desk turns the winning thesis into a real proposal: a clear direction, a disciplined entry, a sized stake and defined risk limits. Every call is logged and held accountable to what the market actually did.
An independent committee, covering market, credit, liquidity, model and operational risk, stress-tests the trade and can veto it. Only trades that clear risk move forward.
Every cleared call crosses the founder's desk for a final read before it reaches a client, exactly as a bulge-bracket desk operates.
After the close, the proprietary Sephiance algorithm learns from how the day's calls played out and adapts the desk, so tomorrow's traders start sharper than today's.
Each pod is a separate strategy with its own written mandate: what it is for, what it trades, its horizon and when it is expected to struggle. Two pods produce live signals the firm’s traders execute; the rest incubate until their evidence clears the bar. How each pod picks its positions is proprietary.
| Pod | Stage | Role | Market | Horizon |
|---|---|---|---|---|
| SephTrader | Live signal | Earn a risk premium | US large-cap equities | Weeks |
| SephWide | Live signal | Earn a risk premium | US large and mid-cap equities | Weeks |
| SephLab | Incubating | Earn a risk premium | Engine check (twin) | Weeks |
| SephLab Wide | Incubating | Earn a risk premium | Engine check (twin) | Weeks |
| SephFactor | Incubating | Predict returns | US large-cap equities | Weeks |
| SephLS | Incubating | Predict returns | US equities, long/short | Weeks |
| SephMacro | Incubating | Protect in drawdowns | Multi-asset | Months |
| SephFX | Incubating | Predict returns | Currencies | Weeks |
| SephCarry | Incubating | Earn a risk premium | Currencies | Months |
| SephPairs | Incubating | Predict returns | US equities, relative value | Weeks |
| SephRegime | Incubating | Protect in drawdowns | Multi-asset | Months |
| SephOptions | Incubating | Earn a risk premium | Equity options | Weeks |
| SephLive | Incubating | Predict returns | US large-cap equities | Weeks |
| SephCrypto | Shadow | Earn a risk premium | Crypto | Weeks |
| SephFast | Incubating | Predict returns | US equities, intraday | Intraday |
| SephTrend | Shadow | Protect in drawdowns | Multi-asset | Months |
| TSMOM pod | Shadow | Protect in drawdowns | Multi-asset | Months |
Each review follows an institutional governance process: divisional analysis, independent risk assessment, structured challenge, and a single accountable decision.
No strategy reaches the desk on a backtest alone.
A dedicated Linux scheduler runs every job on timers that catch up after a missed run.
The AI layer is optional, replaceable and inexpensive. You choose where it runs and what it may read.
Self-hosted on your own server, or on-premise. Uploaded documents are encrypted, isolated to their owner, deleted on request and expire after seven days.
Every AI feature works with a local model (Ollama, or any OpenAI-compatible server), so research and documents never leave your network. Hosted models are optional: DeepSeek, OpenAI, xAI or OpenRouter.
Document review and data-room answers use AI only when you allow it, and private documents bypass every shared cache and log.
Numbers come from deterministic models; the AI explains and challenges them. That keeps prompts short and answers checkable.
Measured on our own usage log: about $0.0015 per model call, and a full investment-committee review costs cents, not dollars. Repeat questions are served from cache at no cost.
Light and deep tasks can run on different providers, and the committee can run fully local. Only the committee reasons at length; every other caller is set to concise answers.
Most market tools, and most AI products, stop at narrative. Sephiance Qant does not. Every number in a report is produced by a deterministic engine built on published, peer-reviewed mathematics, the same toolkit used on institutional quant desks: Monte Carlo simulation, Hidden Markov regime detection, Gaussian Process forecasting, PCA factor decomposition, cointegration testing and more. But the published models are only the foundation. What truly sets us apart is the proprietary Sephiance algorithm that sits above them, tested on more than a decade of market data and continuously adapting the entire engine to real markets. Its inner workings are our own and are not disclosed. The underlying methods are established and reproducible; the way we combine, weight and evolve them is not. Language explains the result. Mathematics produces it.
The published mathematics is only the foundation. Above it sits the Sephiance algorithm, a proprietary self-learning layer developed and tested on more than a decade of market data. It continuously measures the engine against real market behaviour and adapts it, so the system producing tomorrow's research is never quite the same as today's. How it decides what to change, and how the many models are weighted and reconciled into a single call, is the core intellectual property of Sephiance Qant and is deliberately not disclosed. What matters to you is the outcome: every figure you receive is the product of the whole engine, unified and continuously sharpened, never one model's isolated guess.
Deterministic, reproducible, auditable. Narrative explains a call, but it never overrides the mathematics. Where the two disagree, the conflict is flagged and surfaced rather than smoothed over.
Our research runs on a proprietary institutional terminal: a single, live command-line cockpit for cross-asset markets, intelligence, and the full quantitative engine.
Every model, signal and data feed was built for our own capital first. The engine runs the firm’s model portfolios every trading day, recorded since July 2026: SephWide +11.34% since 9 August against +0.94% for the S&P 500, and SephTrader +6.28% since 24 July against +5.73%. It is now open for commercial use.
Real-time futures, world indices, commodities, rates, FX and crypto, alongside global market hours, in a single live view.
A Bloomberg-style interface: type a ticker and a code (AAPL GP, MSFT DES) to drive more than a hundred analysis functions instantly.
Institutional-grade market, news and macro data, unified into a single live view.
The same multidivisional engine that powers every Sephiance Qant report, exposed as a working terminal. Built for our own capital first, now available for commercial deployment. Proprietary, confidential, and built for institutional use.
Everything in the terminal also runs on the web: the same tools, the same data, nothing to install. Screens shown as they appear in the app.
Sephiance Qant does the work of a deal team’s analysts, a fund’s back office, a quant research desk and a compliance function. It is built for private equity firms, family offices and independent managers that cannot justify a six-figure vendor stack.
QoE from a trial balance, journal-entry testing, legal clause review, data-room Q&A, sanctions screening, red-team analysis and IC memos.
LBO, merger accretion/dilution, comps regression, precedents, statement DCF, Monte Carlo DCF and reverse DCF.
Spreading, covenant monitoring, breach probability, distance to default and bank call-report financials.
IRR, TVPI/DPI, PME, carry waterfalls, ILPA LP statements, NAV desmoothing and commitment pacing.
Registries in four jurisdictions, deal radar, government contracts, regulator feeds, filings by severity and global news.
Shipping, air travel, power demand and prices, and degree days: activity read before results are published.
Decision journal, forecaster scoring, model-risk packs, DORA, MiFID, AIFMD and SFDR evidence, and an evidence ledger.
Cross-asset terminal, factor research, portfolio optimisation, stress tests and seventeen strategy pods.
The same toolkit serves very different jobs. Here is what each type of firm uses, and what it gains.
Oversee managers, direct deals and the family's own portfolio without hiring a research team.
The benefit: The diligence and reporting of an institutional office, at the cost of a single subscription.
Source, diligence, value and monitor deals end to end.
The benefit: A first-pass QoE and IC memo in hours, before paying for an outside engagement.
Research, test and run systematic and discretionary strategies with institutional discipline.
The benefit: Stops capital going into strategies that only worked in hindsight.
Value targets, watch competitors and support board decisions.
The benefit: Board-grade analysis without a standing advisory retainer.
Underwrite and monitor borrowers from their own numbers.
The benefit: Earlier warning on a covenant breach, with the evidence on file.
Deliver institutional-quality work to clients as a small team.
The benefit: A research department's output from a one- or two-person firm.
Upload a trial balance, and optionally bank statements, a sales listing and a payroll register. Six stages produce a financial due-diligence report as a PDF plus an Excel databook. Tested against a published sample QoE (QoEPro's fictional Alpine Air HVAC): with the same judgements it matches 9 of the report's 10 headline figures, and the tenth is an inconsistency inside the published report itself. Because that test trial balance was rebuilt from the report, this checks the method rather than independent data.
Every tool is available in the terminal and on the web, and most ship with sample files you can try in a minute.
Buy-side and sell-side work on one company
For lenders: spread a borrower and watch its covenants
Fund performance, carry, LP statements and portfolio companies
Find companies, follow filings, regulators and public money
Read activity before results are published
Who was right, how often, and with what confidence
Model risk, DORA, MiFID, AIFMD, SFDR and markets
17 private equity and M&A functions and 18 valuation and finance models, linked to live market data.
Every tool uses an established method. Missing data is shown as unknown, never as zero.
A chat AI is a brilliant generalist. It is not a system of record: it does not hold your data, prove its inputs, remember what failed, run at night or keep an audit trail. Here is where that difference decides the outcome.
A QoE is only as good as its weakest input cell. These are the places a chat-built QoE goes wrong, and what the workbench does instead.
| What can go wrong | With a chat AI | With Sephiance |
|---|---|---|
| Incomplete data | Computes EBITDA from whatever extract it is given, even one that does not balance. | Proves every period balances (debits = credits) before any figure is used, and measures mapping coverage against a 98% bar. |
| Real-world files | A full general ledger is hundreds of thousands of lines: a chat truncates or summarises it, and misreads "(15,000)", "1.234,56" or 31/12/2024. | Reads trial balances up to 20 MB and listings up to 50 MB line by line: brackets, CR/DR, European decimals and dates, wide or debit/credit layouts. Unreadable cells are listed by cell. |
| Year-to-date ledgers | Treats cumulative balances as monthly, or differences them wrongly across cost centres. | Converts YTD to monthly after summing cost centres, with gaps reported as unknown. |
| Account mapping | Guesses categories silently; an expense can land on the balance sheet and inflate EBITDA. | Maps every account in seven languages with the rule shown, takes your overrides, and lists unmapped accounts instead of hiding them in "other". |
| Who decides an add-back | Makes adjustments itself, and the reasoning changes from run to run. | Detects leads (one-off names, statistical spikes, cut-off, reserve swings, related parties) and moves EBITDA only when a reviewer accepts them. |
| The EBITDA bridge | One adjusted number with no trail. | Reported, management, diligence and pro-forma EBITDA for the LTM and every fiscal year, each step with its basis and source cells, plus a Monte Carlo range. |
| Working-capital peg | An average of whatever months it saw. | LTM average with a bootstrap 95% interval, a seasonal peg by completion month, trend and window-dressing checks, and debt-like items removed so nothing is counted twice. |
| Net debt | Debt plus cash. | Debt-like items found by name in seven languages (accrued interest, income tax, earn-outs, shareholder loans, pensions, provisions, deposits, factoring), restricted cash, and a cash profile through the year. |
| Proof of cash | Not attempted. | Bank receipts against cash-basis revenue month by month, VAT-aware, with each gap flagged. |
| Revenue and payroll quality | Not attempted, or a rough view. | Invoice-listing reconciliation, concentration, cohorts and retention, price-volume-mix, credit notes after year end, duplicates, Benford; payroll headcount, register-to-ledger reconciliation and a run-rate lead. |
| Journal-entry testing | Not possible on a full ledger in a chat. | Benford (first and first-two digits, Nigrini bands), round amounts, weekend and year-end postings, just-below-approval amounts and outliers, ranked into a selection list. |
| The deliverable | Prose in a chat window. | A financial due-diligence report (PDF) and an Excel databook with live formulas that agree with the report, with an open-items list derived from the analysis. |
| Proof it works | None. | Tested against a published sample QoE: 9 of 10 headline figures agree with the same judgements, plus about 5,500 automated tests. |
An AI could help write this code, and parts of it were. What it cannot be is the system that holds the data, proves the inputs, keeps the judgement with a person, runs every night, keeps the audit trail and stands behind its numbers.
Toptal asked its finance consultants which free sources they piece together instead of a $25,000-a-year Bloomberg seat. Here is each one, what it gives, and where the same job is done inside Sephiance Qant.
| Source | What it gives | Status | In Sephiance |
|---|---|---|---|
| Financial modelling & analysis | |||
| Koyfin | Macro dashboards, security snapshots, estimates, charts | Covered | Macro Dashboard, company profiles, consensus, price charts, Markets |
| Atom Finance | Event calendar, valuation from consensus, transcripts, peers | Covered | Earnings calendars, DCF and Monte Carlo DCF, transcript search, comps regression |
| Aswath Damodaran | Industry margins, betas, cost of capital, multiples | Covered | DAMO Industry Benchmarks |
| Finviz | Snapshot, dividends, ratings, price targets, screener | Covered | Quick Scan, consensus, Screener, Watchlist Screener |
| YCharts | FactSet-style data and charts (paid after trial) | Covered | Markets, Macro, charts and fundamentals included |
| Eloquens | Free financial model templates | Covered | XLMODEL editable LBO, DCF and merger workbooks |
| Yahoo Finance | Revenue and EPS estimates, live prices | Covered | Consensus estimates; live market tape |
| SEC EDGAR | 10-Q, S-1 and every public filing | Covered | Event Feed, Filing Cross-check, insider and 13F tools, FORMD |
| IMPORTXML / GOOGLEFINANCE | Data and prices inside spreadsheets | Covered | Live CSV link for any tool in Excel or Google Sheets, plus Excel export |
| Market research | |||
| ABI Research, IDC, Gartner | Paid industry research | Not free | Proprietary; no free equivalent |
| Statista | Industry and consumer statistics | Partly | TAM Market Sizing from official statistics; survey data not replicated |
| MarketsandMarkets, Research and Markets | Niche industry reports | Partly | TAM covers industry size; analyst forecasts not replicated |
| TAM workshop | How to size a market | Covered | TAM tool with official US and EU industry data |
| Macro data | |||
| Trading Economics | Macro by country, with forecasts | Partly | MACROX with IMF projections; their own forecasts not replicated |
| World Bank | Economic and development datasets | Partly | MACROX via DBnomics (IMF World Economic Outlook) |
| BLS | CPI, unemployment, productivity | Covered | Macro Dashboard via FRED; BLS industry data in TAM |
| Investing.com | CDS and bond yields | Partly | Treasury curve, rate anchors, CSPRD spreads by rating; single-name CDS has no free source |
| VC & startup databases | |||
| Fundz, Crunchbase | Funded companies, rounds, founders, investors | Partly | FORMD every US private raise from SEC Form D; investor names and valuations are not public |
| Jay Ritter IPO data | Underpricing, age, VC-backed and dual-class IPO statistics | Covered | IPOSTAT IPO Statistics |
| TechLeap | European startup database | Not free | No free equivalent; UK and Greek registries covered by REG |
| Stock research | |||
| Morningstar | Analyst research reports | Covered | Coverage notes and the 14-agent analysis |
| ValueInvestorsClub | Investment write-ups | Covered | IC Memo and Red Team: the long and the short case |
| Seeking Alpha | Community and sell-side snippets | Partly | Global news, Event Feed, Night Shift; no community content |
Source list: Toptal, “How to Recreate the Resources of a Bloomberg Terminal for Free” (updated July 2024).
What firms usually buy for each capability, what it costs, the closest free or open-source option, and what Sephiance includes. The free options are mostly single libraries or raw data that someone still has to build into a working tool.
| Capability | Usually bought from | Typical price | Free / GitHub alternative | In Sephiance |
|---|---|---|---|---|
| Quality of Earnings | Big-4 or mid-tier FDD engagement | $30k–150k per deal | None. Excel templates only | QOE, JET, QRANGE, WEATHER, REFCHK |
| Valuation, LBO, merger, comps | Capital IQ, PitchBook, Macabacus | $15k–40k per seat/yr | Excel templates; a few DCF notebooks on GitHub | LBO, MERGE, COMPS, PREC, VAL, MCDCF |
| M&A, IPO, activist deal flow | Mergermarket, Dealogic | $20k–50k/yr | SEC EDGAR full-text search (manual) | MA, IPO, SPAC, ACT, INSD, RADAR |
| Legal DD and data-room AI | Kira, Luminance, Datasite AI | $20k–100k/yr | RAG demos with no citation checks | LDD, ROOM, document review |
| Sanctions and counterparty screening | World-Check, Dow Jones Risk, ComplyAdvantage | $5k–50k/yr | OpenSanctions (free for non-commercial use), raw OFAC lists | SCREEN with official lists and registries |
| Credit spreading and covenants | Moody's CreditLens, S&P Capital IQ | $20k–80k/yr | None maintained | SPREAD, COV, COVMC, DTD, BANK |
| Fund performance, waterfall, LP reports | Carta, Allvue, eFront, Juniper Square | $10k–75k/yr | pyxirr, small waterfall scripts | FUND, WATER, LPR, BOOT, DESM, YALE |
| Portfolio-company monitoring | iLEVEL, Chronograph | $15k–50k/yr | Spreadsheets | PMON |
| Company research terminal | Bloomberg Terminal | ~$30k per seat/yr | OpenBB, yfinance | DES, FA, VAL, CONS, OWN, RED, CRED, COVER |
| Fundamentals and screening | FactSet, Refinitiv Workspace | $15k–30k/yr | OpenBB, Finviz free tier, edgartools | FILE, BASE, factor screens |
| Alternative data | Kpler, Vortexa, Quiver | $10k–100k+/yr | Raw public feeds you wire up yourself | SHIP, NOW, NOAA, GOV, ATTN, GDELT |
| Risk models and stress | MSCI Barra, Axioma, RiskMetrics | $25k–80k/yr | Riskfolio-Lib, PyPortfolioOpt, QuantStats | STRESS, SCEN, cross-portfolio risk |
| Quant research and backtesting | FactSet Alpha Testing, QuantConnect, CRSP data | $5k–40k/yr | LEAN, Backtrader, Zipline, vectorbt, Qlib | Validated backtests, nightly research, point-in-time data |
| Model risk and regulatory evidence | Consultants, regulatory reporting vendors | $10k–50k/yr | None | MRM, COMPLY, LEDGER, REGS |
| Country macro for any country | Trading Economics, Bloomberg ECO | $49/week pass to Bloomberg seat | World Bank / IMF downloads by hand | MACROX |
| Industry benchmarks | Capital IQ, FactSet industry data | $15k+/yr | Damodaran spreadsheets by hand | DAMO |
| Credit spreads | Bloomberg CDS screens, Markit | Bloomberg seat | Investing.com pages | CSPRD |
| Private funding rounds | Crunchbase Pro, PitchBook, Fundz | $500–20k+/yr | Crunchbase free tier | FORMD |
| IPO base rates | Dealogic, Renaissance Capital | $10k+/yr | Ritter PDFs | IPOSTAT |
| Market sizing | Statista, IBISWorld, MarketsandMarkets | $1k–10k per report | Census / Eurostat tables by hand | TAM |
| Model templates | Macabacus, Eloquens | $300+/yr | Static Excel templates | XLMODEL |
Prices are typical public list prices or quotes, for orientation only.
A full research desk, on demand. Institutional-grade quantitative analysis across asset classes.
Deep equity research, portfolio construction and optimisation to grow and protect your holdings.
Equities & portfoliosCross-asset coverage of commodities, FX, crypto, rates and the Greek market, with live macro intelligence.
Cross-asset coverageIndependent backtesting, factor screening and quantitative validation before any call is released.
Validation & oversightThe most complete personal finance app for the Eurozone. #1 in Greece with 15,000+ users.
The tools a large institution pays six figures for, from €190 a month. Billed monthly, cancel anytime.
Market information and research highlights, straight to your inbox.
For individual investors: research, market notes and the Sephiance app.
One seat for an independent manager or advisor. The full terminal and web platform.
For private equity firms and family offices, up to five seats.
On-premise deployment for larger firms, on your own data licences.
For firms without a subscription. Priced by deal size and scope.
The daily buy and sell signals from the SephTrader and SephWide algorithmic books.
Signal returns are recorded book returns over a short period and are not a guarantee of future results. Prices exclude VAT.
Architect of Sephiance Qant. Designed and built the entire platform: the four-division research structure, the quantitative engine, and more than a hundred analyst tools covering deal diligence, quality of earnings, valuation, credit, fund reporting and compliance. Every result is tested on point-in-time data and checked by independent code before it counts, so the platform says what it knows and flags what it doesn’t.
Built by a team with deep roots in quantitative finance and software engineering. The terminal delivers the calibre of analysis expected on a sell-side trading desk, powered by a proprietary database built over years and a purpose-built quantitative engine.
"Everyone should have access to a high-end terminal and the data they need to make informed financial decisions, without having to go through gatekeepers."
Ioannis Tzalas
Join over 10,000 serious investors who trust Ioannis Tzalas for a clear read on the Fed, inflation, and where markets are heading. No hype. Just the calls that actually move portfolios.
Free forever. Unsubscribe anytime.
Deal diligence, credit, fund reporting, sourcing, nowcasting and compliance tools, all in the terminal and on the web. New this month: country macro for any country, credit spreads, private raises from Form D, IPO statistics, market sizing, industry benchmarks and editable Excel models.
Upload a trial balance and get a Big-4 style QoE: adjusted EBITDA bridge, working-capital peg, net debt, proof of cash, and revenue and payroll quality, delivered as a PDF report with an Excel databook.
To 9 October, SephWide is up 11.34% since 9 August against 0.94% for the S&P 500, and SephTrader is up 6.28% since 24 July against 5.73%. Daily signals from both books are now available by subscription.
Full Greek and EU data protection compliance. Your data stays yours.
All transactions processed through Stripe. We never store card data.
Our mobile app collects nothing. Everything stays on your device.
Operated from Athens under Greek law and EU regulation.